The disconnect here for me is, I assume the DoW and Anthropic signed a contract at some point and that contract most likely stipulated that these are the things they can do and these are the things they can't do.
I would assume the original terms the DoW is now railing against were in those original contracts that they signed. In that case it looks like the DoW is acting in bad faith here, they signed the original contact and agreed to those terms, then they went back and said no, you need to remove those safeguards to which Anthropic is (rightly so) saying no.
Am I missing something here?
EDIT: Re-reading Dario's post[1] from this morning I'm not missing anything. Those use cases were never part of the original contacts:
> Two such use cases have never been included in our contracts with the Department of War
So yeah this seems pretty cut and dry. Dow signed a contract with Anthropic and agreed to those terms. Then they decided to go back and renege on those original terms to which Anthropic said no. Then they promptly threw a temper tantrum on social media and designated them as a supply chain risk as retaliation.
My final opinion on this is Dario and Anthropic is in the right and the DoW is acting in bad faith by trying to alter the terms of their original contracts. And this doesn't even take into consideration the moral and ethical implications.
> *Isn’t it unreasonable for Anthropic to suddenly set terms in their contract?* The terms were in the original contract, which the Pentagon agreed to. It’s the Pentagon who’s trying to break the original contract and unilaterally change the terms, not Anthropic.
> *Doesn’t the Pentagon have a right to sign or not sign any contract they choose?* Yes. Anthropic is the one saying that the Pentagon shouldn’t work with them if it doesn’t want to. The Pentagon is the one trying to force Anthropic to sign the new contract.
I just wish there was a stronger source on this. I am inclined to agree you and the source you cited, but unfortunately
> [1] This story requires some reading between the lines - the exact text of the contract isn’t available - but something like it is suggested by the way both sides have been presenting the negotiations.
I deal with far too many people who won't believe me without 10 bullet-proof sources but get very angry with me if I won't take their word without a source :(
I think a big question mark here, is whether anything said on Anthropic's side if in the framing of "We have a thing going on that we are trying to communicate around where a canary notice if it existed would no longer be updated"
Also, Trump's own words complaining about being forced to stick to Anthropic's terms of service:
> The Leftwing nut jobs at Anthropic have made a DISASTROUS MISTAKE trying to STRONG-ARM the Department of War, and force them to obey their Terms of Service instead of our Constitution.
It isn't about commercial agreements, it's about patriotism. The national industry is supposed to submit to the military's wishes to the extent that they get compensated. Here it's a question or virtue.
The Pentagon feels it isn't Anthropic to set boundaries as to how their tech is used (for defense) since it can't force its will, then it bans doing business with them.
No one cares if the Pentagon refuses to do business with Anthropic. But Hegseth has declared that effective immediately, no one else working with the DoD can either--which includes the companies hosting Anthropics models (Amazon, Microsoft, and Alphabet).
So it's six months to phase out use of Anthropic at the DoD, but the people hosting the models have to stop "immediately".
Which miiight impact the amount of inference the DoD would be able to get done in those six months.
Regardless of the original contract, it's entirely appropriate for a vendor to tell the customer how to use any materials.
Imagine a _leaded_ pipe supplier not being allowed to tell the department of war they shouldn't use leaded pipes for drinking water! It's the job of the vendor to tell the customer appropriate usage.
Playing devil's advocate: if I did in fact grab one of my kitchen knives to defend myself against a violent intruder into my kitchen, I wouldn't expect to be banned from buying kitchen knives.
I'm not sure this is still a useful analogy, though...
If I shoot someone, something that is explicitly warned against in firearm safety materials that come with every purchase of a new firearm, I am no longer allowed to purchase any more firearms.
And if you grabbed the knife and went on a violent spree, I'd absolutely expect the knife manufacturer to refuse to sell to you anymore.
The knife manufacturer isn't obligated to sell to you in either case, I'd expect them not to cut ties with you in the self defence scenario. But it is their choice.
The specific shape of a kitchen knife would make it a particularly poor fighting knife, and knives in general are bad for self defense, due to the potential for it to be turned against the user. So, there is a good argument that such a suggestion is really in the user's best interest rather than a cynical play for the manufacturer to limit liability.
Not in software though. Clear precedent has been established via EULAs. Software companies set the rules and if users don't like, they can piss off. I don't see why it would be any different for the government.
I'm not a fan of EULAs, I think if you acquire some software anonymously and run it on your own systems you should be able to do whatever you want. however if you want software hosted on someone else's machines, or want to enter into a contractual relationship with them then government or not you should not have the right to compel work from them.
The government is armed and can exempt itself from prosecution either by judicial means and/or by naked force. So it isn’t just a cut and dry licensing problem.
The government cannot set arbitrary rules, it has to follow the law. (And, at least with a functioning separation of powers, it cannot change the law arbitrarily.)
Claude Opus is just remarkably good at analysis IMO, much better than any competitor I’ve tried. It was remarkably good and complete at helping me with some health issues I’ve had in the past few months. If you were to turn that kind of analytical power in a way to observe the behaviour of American citizens and to change it perhaps, to make them vote a certain way. Or something like - finding terrorists, finding patterns that help you identify undocumented people.
Yes, it's officially still the Department of Defense.
If this were a news outline writing "Department of War" I would be concerned. But in the case of the Anthropic CEO's blog post, I can understand why they are picking their fights.
They can, however, rename their Twitter/X accounts and vacate the @SecDef handle, which seems to be up for grabs now, if anyone wants to do the funniest thing...
Of all the silly things that Trump did, I think this one is the most reasonable. This has always been a department of war. Calling it defense was propaganda.
It's a silly shibboleth, but I automatically ignore anyone who calls it the Department of War or Gulf of America. Hasn't steered me wrong yet. They're telling me they're the kind of people who only care about defending fascism.
I think it's worth giving people a tiny bit of grace on this. I've surprised people by explaining that the "Department of War" is just fascist fanfic and that the legal name has not changed.
It's a testament to the broken information ecosystem we're in that many people genuinely don't know this. Most will correct themselves when told. I agree with you that those who don't are not worth engaging.
I would not defend all of Google's decisions in the Trump era, but complying immediately with politicized name changes has always been the status quo. Even in healthy democracies, the precise names of geographic features can be extremely controversial, and no sane company wants to get in a debate with the Japanese government about the real names of various islands.
the entire administration negotiates in bad faith. literally every agreement they sign whether it's international trade or corporate contracts is up to the whim of a toddler with twitter
With this administration, after all their proven lies, when in doubt, assume bad faith on their part. Assuming good faith at this point is Lucy and Charlie Brown and the football, but now the football is fascism (i.e., state control of corporations, e.g., what Trump administration is doing here).
Trump has historically stiffed his contractors. Why do you think his administration would be any different with adhering to a contract?
I was pondering the same thing and to me the answer is a contractor sold something to the DoD and Anthropic pulled the rug out from under that contractor and the DoD isn't happy about losing that.
My speculation is the "business records" domestic surveillance loophole Bush expanded (and that Palantir is build to service). That's usually how the government double-speaks its very real domestic surveillance programs. "It's technically not the government spying on you, it's private companies!" It's also why Hegseth can claim Anthropic is lying. It's not about direct government contracts. It's about contractors and the business records funnel.
Yes, I assumed a mass surveillance Palantir program also. Interesting take on how it allows them to claim “we are not doing this” while asking Anthropic to do it.
Of course they can just say - we aren’t, Palantir is.
Wow, and the only restrictions Anthropic asked for are (1) no mass domestic surveillance and (2) require human-in-the-loop for killing [1]. Those seem exceptionally reasonable, and even rather weak, lol :|
Anthropic specifically called out systems "that take humans out of the loop entirely and automate selecting and engaging targets".
I take that to mean they don't want the military using Claude to decide who to kill. As a hyperbolic yet frankly realistic example, they don't want Claude to make a mistake and direct the military to kill innocent children accidentally identified as narco-terrorists.
At least, that's the most charitable interpretation of everything going on. I suspect they are also worried that the sitting administration wants to use AI to help them execute a full autocratic takeover of the United States, so they're attempting to kill one of the world's most innovative companies to set an example and pressure other AI labs into letting their technology be used for such purposes.
Their intention is to turn it against the American people. Hegseth literally wrote a book about eliminating democrats from the US, and this surprises people.
So they are such a risk to national security that no contractor that works with the federal government may use them, but they're going to keep using them for six more months? So I guess our national security is significantly at risk for the next six months?
SCOTUS says POTUS is above the law, so POTUS has collected $4B in bribe / protection money since taking office 13 months ago. Anthropic has lots of money at the moment. Why should they be allow to keep it?
Since they didn't pay off the president (enough?), his goons are going to screw with their revenue and run a PR smear campaign.
Once you realize it only has to do with Trump's personal finances, and nothing to do with national security or the rule of law, then all the administration's actions make perfect rational sense.
Open question: How much should a congress-critter charge Trump for a favorable vote? (The check should come with a presidential pardon in the envelope, of course...)
They are the same amount of ‘risk’ to national security that the various ‘emergencies’ the executive branch has used as legal excuses to do otherwise illegal things are emergencies.
Congress is negligent in not reigning this kind of thing in. We’re rapidly falling down so many slippery semantic slopes.
The Trump administration tends to use this playbook.
Putting aside my take, I’m trying to objectively make sure I’m grounded on what is likely to happen next, without confusing “what is” with “what is ok”.
Don't make the mistake of thinking their words have meaning. They see a way to punish the company, they take it. Same thing with declaring a national emergency to impose tariffs. There's no supply chain risk, no national emergency, but that doesn't stop them.
Dont forget Nvidia technology was condsidered too sensitive to be exported to China....until the Trump administration decided they could export it if they paid a 10% export tax.
> So I guess our national security is significantly at risk for the next six months?
That does seem to be what Hegseth is arguing, yes; and that is presumably his justification for doing something drastic here. Although I assume he is lying or wrong.
And as a cynic, let me just add that the image of someone going to the political overseers of the US military with arguments about being "effective" or "altruistic" is just hilarious given their history over the last ~40 years.
It's also a very clear differentiator for them relative to Google, Facebook, and OpenAI, all of whom are clearly varying degrees of willing to sell themselves out for evil purposes.
It will also cost openai dearly if they don't communicate clearly, because I for one will internally push to switch from openai (we are on azure actually) to anthropic. Besides that my private account also.
Given the history of US military adventurism and that we’re about to start another completely unjustified war of aggression against Iran, yes. Absolutely yes.
> Costco is a really popular subject for business-success case studies but I feel like business guys kinda lose interest when the upshot of the study is like "just operate with scrupulous integrity in all facets and levels of your business for four decades" and not some easy-to-fix gimmick
I don't know, staff at my two Costcos feel much more disinterested and rude then I remember a decade ago. It used to feel fun but now it's miserable.
At peak times they run out of carts and tell the customers to go hunting in the lot for them, door greeters shouting at members across the floor, checkout queues stretch the length of the warehouse, they start half blocking the gas station entrance 30mins before close so trucks can't get in, so maybe they're turning those profit screws.
Ah, right, by being actually good, as in - being okay with mass surveillance as long as it isn't being done in the US, being okay with Claude assisting in killing people as long as it isn't fully autonomous, and being actively hostile to open-weight LLMs and open research on LLMs? This kind of "good"?
No, OP is right, their PR department is doing a great job.
Correct. Protect our citizens' rights, as we are the ones under the jurisdiction of our government. Yes, design competitive weapons systems that can stand up to the threats that adversary powers are creating, but do so while maintaining human control.
This whole saga is extremely depressing and dystopic.
Anthropic is holding firm on incredibly weak red lines. No mass surveillance for Americans, ok for everyone else, and ok to automatic war machines, just not fully unmanned until they can guarantee a certain quality.
This should be a laughably spineless position. But under this administration it is taken as an affront to the president and results in the government lashing out.
If you're a billionaire there's no risk to "sticking to principles", so there's nothing to admire. Also that's not what they're doing. These are calculated moves in a negotiation and the trump regime only has 3 years left. Even a CEO can think 4 years ahead.
It's probably in Anthropic's interest to throw grok to these clowns and watch them fail to build anything with it for 3 years.
i disagree. 3 years is an insanely long time in the AI space. The entire industry pretty much didn't even exist three years ago! Or at least not within 4 orders of magnitude.
Also, every other company has bent the knee and kissed the ring. And the trump admin will absolutely do everything they can to not appear weak and harm Anthropic. If it was so easy to act principled, don't you think other companies would've refused too? Eg Apple
And there is real harm here. You're reading about it - they get labeled a supply chain risk. This is negative and very tangible
Good. I'd rather not have my favorite AI from a company working on AGI to have murder and spying in it's DNA.
In fact, as a patriotic American veteran, I'd be ok with Anthropic moving to Europe. It might be better for Claude and AGI, which are overriding issues for me.
Rutger Bregman @rcbregman
This is a huge opportunity for Europe. Welcome Anthropic with open arms. Roll out the red carpet. Visa for all employees.
Europe already controls the AI hardware bottleneck through ASML. Add the world's leading AI safety lab and you have the foundations of an AI superpower.
> Good. I'd rather not have my favorite AI from a company working on AGI to have murder and spying in it's DNA.
Anthropic made it quite clear they are cool with spying in general, just not domestic spying on Americans, and the "no killbots" pledge was asterisked with "because we don't believe the technology is reliable enough for those stakes yet". The implication being that they absolutely would do killbots once they think they can nail the execution (pun intended).
I suppose you could say they're taking the high road relative to their peers, but that's an extremely low bar.
I wouldn't say it's clear. People keep pointing to the wording used in the statement to say it, but I wonder if it has to do with constitutionally; domestic surveillance of people in the US without a warrant is against the constitution, and surveillance of non-citizens outside the U.S is not. Can they even be compelled by the executive branch to do an action that may be unconstitutional?
Do all of the employees want to move to Europe suddenly? Unless it’s the UK or Ireland, do they speak the local language? If it is the UK or Ireland, do they prefer the weather in California? Do they have children in school or in college locally? Do they have family they’d rather not move 9 time zones away from? Elderly parents they’re taking care of?
I'm pretty vocal about our collective authority to work against the Trump administration, and even I would be hesitant to work as a US employee of a company that fled the country after a dispute with the US military. Seems like an extreme threat to my personal safety for little resistance benefit.
Canada is another option. Canada has significant AI research institutes going back decades ( https://en.wikipedia.org/wiki/Mila_(research_institute) ) that have produced much of the foundational research that backs today's AI models.
For Americans and international researchers it's easy to get visas there quickly. It's not far at all for Americans to relocate to or visit. Electricity is cheap and clean. Canada has the most college educated adults per capita. The country's commitment to liberalism, and free markets, is also seeming more steadfast than the US at this point in time.
Canada faces obstacles with its much smaller VC ecosystem, its smaller domestic market, and the threat of US economic aggression. Canada's recent trade deals are likely to help there.
I say this all as an American who is loyal to American values first and foremost. If the US wants to move away from its core values I hope other countries, like Canada or the EU, can carry on as successful examples for the US to eventually return to.
I have my doubts about Anthropic wanting to pick up and move the entire company to Europe even if Ursula von der Leyen personally signed their visas. Maybe only if the government tried to nationalise their proprietary models.
Why wouldn’t the government just arrest their board and execs on charges of treason or something? At this point they could probably publicly hang them all and a plurality of Americans would cheer it. I don’t know if you appreciate how disliked tech is by the left and right alike.
The left would never support that lawlessness: opposition to AI is based on things like ethics, environmental impact, etc. which are predicated on concepts like the rule of law. People are calling for regulation or UBI, mor killings.
The right has far more talk of violence, true, but a lot of that is targeted rhetoric to keep voters riled up, and it’s not aimed at American businesses. I’d be surprised if even a third of Republicans supported anything more than not doing business with Anthropic. Even the Nvidia shakedown got a ton of criticism and that’s just money.
"I can state flatly that heavier than air flying machines are impossible.
— Lord Kelvin, 1895"
I'm sure this doesn't apply to you since you're not Lord Kelvin. On the other hand, people like Peter Norvig state in a popular AI textbook that, for example, they don't know why similar concepts appear close by in the vector space, so maybe you just know something other people don't.
I'm not taking a position here but the person you're replying to stated that Anthropic are working on AGI, not that their current LLM offering will evolve into AGI.
False, and you've given no argument to the contrary. There's certainly no definition that precludes it. It isn't, currently; there's no reason it can't be, any more than there's reason that Conway's Game of Life can't be, given sufficiently interesting data to process. Any Turing-complete system could simulate AGI. It might not be the most efficient mechanism for doing so, but that's not the question at hand.
Europe doesn’t give a shit about another American company and their employees trying to dominate their markets and import their workaholic American culture. They will tell Anthropic to go home.
"Europe" is not a single entity with uniform opinions. As an European, I would much rather have hardworking people and """workaholic""" culture than regress to an underdeveloped culture fueled by laziness.
The designation says any contractor, supplier, or partner doing business with the US military can’t conduct any commercial activity with Anthropic. Well, AWS has JWCC. Microsoft has Azure Government. Google has DoD contracts. If that language is enforced broadly, then Claude gets kicked off Bedrock, Vertex, and potentially Azure… which is where all the enterprise revenue lives. Claude cannot survive on $200/mo individual powerusers. The math just doesn’t math.
None of the hyper scalers are going to stop offering Claude. All of the big 3 have invested billions of dollars into Anthropic, and have tens (if not hundreds) of billions more tied up in funding deals with them. Amazon and Google are two of the largest shareholders of Anthropic.
Anthropic is going to be fine. The DoD is going to walk this back and pretend it never happened to save face.
Tens, maybe hundreds, of billions? That’s cute. The DoD will spend $961b this year. It does that like clockwork every year, year after year.
Anthropic is not even close to too big to fail. And even if this could get settled in court 5 years from now, this can easily throw enough of a wrench into their revenue streams to kill their flywheel.
That $961 billion includes things like airplanes and bullets, tech companies are only getting a taste of that pie not anywhere close to the whole thing.
Half of that budget gets contracted out to Lockheed, Raytheon, Northrop, Boeing, General Dynamics, etc. Those companies absolutely do spend money on the hyperscalers.
GovCloud revenue is in the tens of billions of dollars. Bedrock less so. Almost every FedRAMP product uses the same codebase for Fed and non-Fed, and this would force most FedRAMP vendors to blackball Anthropic.
The JWCC, which is larger than GovCloud, was only $9b, split across three companies, over ten years. It’s peanuts compared to the investments that the hyperscalers have with Anthropic.
JWCC is not the only project. Vendors like Crowdstrike also rely on hyperscalers to serve their products to federal customers, and the codebase is shared.
This announcement has made Anthropic toxic in the entire dependency chain because it means years of efforts and tens to hundreds of millions of dollars rearchitecting entire platforms and renegotiating contracts.
The entire cybersecurity industry has a TAM of $208 BILLION [0]
> because it means years of efforts and tens to hundreds of millions of dollars rearchitecting entire platforms and renegotiating contracts.
This is exactly why this announcement has not made Anthropic toxic. The entire industry knows how ridiculous this move is from Hegseth, and it’s going to be rolled back next week once the adults get back from their weekend.
I would find that a lot more plausible if people had not spent the past week giving me similar arguments, in precisely the same tone, for why this was an empty threat and would never happen in the first place. If Amazon and Google do not either bow down or immediately join a business coalition to get Trump out of power, Hegseth will be even happier to get an opportunity to prove his power by destroying them. Trump either doesn't want to stop him or has become too senile to stop him.
" Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic."
Example: Perhaps "Amazon US Services LLC" or whichever subsidiary they have that deals with the government will be banned from using Claude, and all of it's other subsidiaries won't?
Well, IANAL but tweets aren't legislation. What that tweet implies is something that would have to be amended into the NDAA, which requires congress. Hegseth can't just go on a drunk rant and have everything out of his mouth become law.
The supply chain risk directive would come from existing procurement law, which only allows the DoD to require contractors to certify that Anthropic is not used in the fulfillment of any government work.
Which is also separate from Trumps' EO, which being an EO only applies to the federal government directly.
So yeah, banning any contractor, supplier, or partner from any commercial activity with Anthropic is just fantasy without going through congress first.
> Contractors can still use Claude internally in their business, so long as it is not used in government work directly.
I work in the enterprise SaaS and cybersecurity industry. There is no way to guarantee that amongst any FedRAMP vendor (which is almost every cybersecurity and enterprise SaaS or on their roadmap).
Almost all FedRAMP products I've built, launched, sold, or funded were the same build as the commerical offering, but with siloed data and network access.
This means the entire security and enterprise SaaS industry will have to shift away from Anthropic unless the DPA is invoked and management is changed.
More likely, I think the DoD/DoW and their vendors will force Anthropic to retrain a sovereign model specifically for the US Gov.
Edit: Can't reply
> This is the core assertion that is not clear nor absolute.
If Walmart can forcibly add verbiage banning AWS from it's vendors and suppliers, the US government absolutely can. At least with Walmart they will accept a segmented environment using GCP+Azure+OCI. Retraining a foundational model to be Gov compliant is a project that would cost billions.
By declaring Anthropic a supply chain risk, it will now be contractually added by everyone becuase no GRC team will allow Anthropic anywhere in a company that even remotely touches FedRAMP and it will be forcibly added into contracts.
No one can guarantee that your codebase was not touched by Claude or a product using Claude in the background, so this will be added contractually.
> This means the entire security and enterprise SaaS industry will have to shift away from Anthropic unless the DPA is invoked and management is changed.
This is the core assertion that is not clear nor absolute.
From what I’ve heard the actual restriction is just on using Claude for stuff they’re doing for the Pentagon. They’ll keep using Claude for everything else and be less effective when they work for the government, and that’s fine because everyone else working for the government will have the same handicap.
This will likely go to court, again as Dario has stated this is blatant retaliation as no US company has ever been designated a supply chain risk and they continue to operate on classified systems for 6 more months.
Yea strong odds this goes to court, the DoD’s clearly inconsistent logic is ridiculed by a judge, the designation is dropped, and everyone quietly goes about their way with the DoD continuing to use Claude according to the existing terms of the contract.
Topics like this are where I struggle with HN philosophy. Normally avoiding politics and ideology where possible, created higher quality and more interesting discussions.
But how do you even begin to discuss that Tweet or this topic without talking about ideology and to contextualize this with other seemingly unrelated things currently going on in the US?
I genuinely don't think I'm conversationally agile enough to both discuss this topic while still able to avoid the political/ideological rabbit-hole.
You can't discuss this topic without broaching the idea that the government is acting in bad faith — that they don't actually believe that Anthropic is a supply-chain risk and that this action is meant to punish the company. But this is in the HN guidelines regarding comments:
> Please respond to the strongest plausible interpretation of what someone says, not a weaker one that's easier to criticize. Assume good faith.
If a commenter who supports the government makes the same argument that the government is making, the guidelines tell us to assume good faith.
My conclusion is that any topic where a commenter might be making a bad faith argument is outside the scope of Hacker News.
My interpretation of that is that I’m required to assume good faith on behalf of other commenters. So, if someone makes the same argument as the government, I’m supposed to assume good faith there, but nothing requires me to assume good faith on behalf of the government. So I can say that this is obviously a shakedown without breaking the rules.
I've been on hn for years and I see this kind of sentiment raised all the time. It is not my understanding of the guidelines.
Politics and ideology are not off topic, provided the subject matter is of interest, or "gratifying", to colleagues in the tech/start-up space.
What's important is that we don't use rhetoric, bad faith or argumentation to force our views on others. But expressing our opinions about how policy affects technology and vice versa has always been welcome, in my observation.
So, what do you think about the US government's decision, and why?
Being a hacker used to be an extremely political and ideological movement. Then capitalism came along and bought the term. It's about time we take that word back where it belongs.
If the last ten years have taught us anything it's that politics just isn't a topic isolated to the halls of government. It's real life. Political alignment has never so starkly indicative of your position on fundamental human morality. At the same time we've never had a government be so directly involved in private businesses.
>Topics like this are where I struggle with HN philosophy. Normally avoiding politics and ideology where possible, created higher quality and more interesting discussions.
Our whole society runs on technology. All tech is inherently political.
A "no politics" stance is merely an endorsement of the status quo.
Why would you want to be non-political in 2026? The current administration is awful in ways we couldn't have imagined. There's no sense in not talking about it.
I appreciate your restraint, and keeping this a high quality discussion space. As a political dissident myself, I don't mind some threads going political, I expect them to. The best ones are when there is a lot of disagreement or debate. As long as its not in every unrelated thread....
In theory, this is why there should be competition in industry, because it removes the capability of a single large actor to be able to control the government's access to things.
Oddly, though, it seems like that should solve this problem as well. I'm not sure why the Department of Defense insists on Anthropic's models in particular; one would think one of the other players, at the very least least xAI, would be willing to step in and provide the capability Anthropic doesn't want to provide.
Everyone is getting wrapped around the axel here but this is about the big picture, not the specifics. A private company should not have the ability to dictate how its technology is used by the government. If they can’t agree to that, then don’t sell your technology to the government. Personally, I don’t want to be spied on by the government with it (I don’t think their tech does that) but I also don’t want Anthropic having operational control over a mission.
McCarthyism began in 1947, with Truman demanding goverment employees be "screened for loyalty". They wanted to remove anyone who was a member of an "organization" they didn't like. It began with hearings, and then blacklists, and then arrests and prison sentences. It lasted until 1959. (https://en.wikipedia.org/wiki/McCarthyism)
This is the new McCarthyism. Do what the administration says, or you will be blacklisted, or worse.
> Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic.
I’m sure the lawyers just got paged, but does this mean the hyperscalers (AWS, GCP) can’t resell Claude anymore to US companies that aren’t doing business with the DoD? That’s rough.
Probably yes. Additionally the (probably more for AWS) won't be allowed to use it internally either. This will probably apply to all the top SaaS/software companies unilaterally.
Additionally, every major university will undoubtedly have to terminate the use of Claude. First on the list will be universities that run labs under DOD contracts (e.g. MIT, Princeton, JHU), DOE contracts (Stanford, University of California, UChicago, Texas A&M, etc...), NSF facilities (UIUC, Arizona, CMU/Pitt, Purdue), NASA (Caltech).
Following that it will be just those who accept DOD/DOE/NSF grants.
Billable hours will win figuring it out but in theory, no because they can’t test it or use it.
Generally any machine that touches Supply chain Risk software cannot ship any software to DoD. AWS has separate clouds but software comes from same place.
AWS/GCP/Azure all do business with the DoD and at least AWS and Azure use Claude a decent amount internally. AWS’s Kiro tool (which is used internally instead of Claude Code) relies entirely on Claude models.
This is almost certainly going to be rolled back, because I guarantee the DoD isn’t going to stop doing business with the hyper scalers, and the hyper scalers aren’t going to stop doing business with Anthropic.
I don't think he got it backwards, at least if Hegseth's statement is accurate. AWS, GCP, etc. all do business with DoD. If they, as DoD contractors, are no longer allowed to do business with Anthropic, then presumably they have to stop re-selling or hosting Anthropic's models to anyone.
Oh you tender babes, trying to logic the meaning of what the lieutenant of the biggest crime syndicate in the world means with his words, as if this was a well thought-out strategy... it's a shakedown; it would make more sense to ask "at least if Hegseth is sober..."
Agree with other reply. I don’t think it’s backward. No they said any commercial activity. Does not feel like a stretch that commercial activity includes reselling api usage.
I don't see how you get that reading. Anthropic is clearly allowed to sell Claude to companies not doing business with the US Military. If anything that's more likely to be non-US companies.
IIRC, the supply chain risk designation is sticky which is why it tends to ultimately mean "nobody can work with this". Amazon using claude means a DoD company can't use Amazon. Every business that touches claude gets tainted.
It's a bit like how the US Cuba sanctions worked and why they effectively isolated Cuba from everything.
Because Anthropic sells Claude through other companies that in turn do business both with Anthropic and the government. These intermediaries, large cloud companies, can't offer Claude anymore if they want to keep the government as a customer.
The government is faaaaaaaaaaaar too invested in Azure and AWS for Microsoft or Amazon to give even half a shit. The DOD has no where else to go and the companies know it. They'll sit on their hands until the legal maneuvers play out, which will take longer than this administration will be in office.
You expect hyperscalers to play chicken with the DoD?
The courts have historically been pretty consistent about giving the DoD whatever the fuck they want, going back to WW2 and even longer for the predecessors of the DoD. I agree that the next administration might reverse it, but the thing is, the government will stay irrational longer than Anthropic will remain solvent.
The US government told every American company to stop doing business with Huawei and they all did it overnight, even when it cost them billions. TSMC stopped fabricating for them, Google pulled Android licensing… The machinery of sanctions compliance is extremely well-oiled and companies fold instantly because the outcome of noncompliance is literally getting thrown in prison.
This is also true, unless the government can force them to drop Anthropic on the basis that the alternative- the government dropping them- is unworkable.
Or Pete Hegseth will threaten to do the same to them unless they comply, and they will demonstrate the same inexcusable cowardice the American business class has consistently demonstrated this past year. Hope I'm wrong and this has finally woken them up!
The stock market will be spooked if the US govt can willy nilly high trajectory darling of the AI world like this though.
Who's next? OpenAI? Google? What if they refuse to allow the DoD to use AI with zero safeguards and Trump's goons decide they are also a "supply chain risk"?
No. The stock market has understood for generations that it's the guys with the guns that protect their gold. The stock market will have a sigh of relief.
"They have threatened to remove us from their systems if we maintain these safeguards; they have also threatened to designate us a “supply chain risk”—a label reserved for US adversaries, never before applied to an American company—and to invoke the Defense Production Act to force the safeguards’ removal. These latter two threats are inherently contradictory: one labels us a security risk; the other labels Claude as essential to national security." from Dario's statement (https://www.anthropic.com/news/statement-department-of-war)
Supply chain risk ? Seems the risk here is the US Gov't wanting free reign to do whatever they want - - when they want.
Look no further than the famous expose by Mark Klein, the former AT&T technician and whistleblower who exposed the NSA's mass surveillance program in 2006, revealing the existence of "Room 641A" in San Francisco. He discovered that AT&T was using a "splitter" to copy and divert internet traffic to the NSA, proving the government was monitoring massive amounts of domestic communication.
If you read Anthropic statement carefully, they explicitly confirm they are already working with the U.S. government on a range of military and national security use cases, many including areas that clearly relate to real world lethal operations.
They are only refusing two narrow, but important categories. Framing this as blanket "refusal to support the DoD" feels like an angry, reactive own goal rather than a careful reading of what they actually said.
So far the march toward dictatorship keep being detoured by sheer incompetence. In any case, is hard to seize power when you can’t organize a group chat...
Basically now all those projects are screwed and need to restart with another provider. I'm sure that's not going to be a massive PITA and delay for all involved.
The military already has access to Grok, but doesn't want it, because it's an inferior model, even compared to open source ones. So the military would probably choose to replace supply chain risk Claude with Qwen or Kimi before Grok.
It would be untouchable irony for the US to cut all ties with Anthropic and replace them with models developed by Chinese labs. The Onion becomes more irrelevant with each passing day.
How many generations does it take before the historians/archeologists uncover old issues of The Onion and decide it was the authoritative news of the day?
Grok is according to most benchmarks pretty close to SOTA. It is where the leaders were just a few weeks ago.
Which exactly is best changes on almost a weekly basis as different companies tweak their best model. I doubt the military would want to be switching supplier every week.
From the public comments over the last few days, my guess is they want a militarized version of Claude. Starting with a box they want to put in the basement of the Pentagon where Antropic can't just switch off the ai. Then some guardrails are probably quite bothersome for the military and they want them removed. Concretely if you try to vibe-target your ICBMs Claude is hopefully telling you that that's a bad idea.
Now, my guess is in the ensuing lawsuit Antropic's defense will be that that is just not a product they offer, somewhat akin to ordering Ford to build a tank variant of the F150.
> Concretely if you try to vibe-target your ICBMs Claude is hopefully telling you that that's a bad idea.
On the non-nuclear battlefield, I expect that the goverment wants Claude to green-light attacks on targets that may actually be non-combatants. Such targets might be military but with a risk of being civilian, or they could be civilians that the government wants to target but can't legally attack.
Humans in the loop would get court-martialed or accused of war crimes for making such targeting calls. But by delegating to AI, the government gets to achieve their policy goals while avoiding having any humans be held accountable for them.
Why can't Grok achieve this? Everyone is saying they don't want to work with Grok because Grok sucks, but it's good enough for generating plausible deniability, isn't it?
> Starting with a box they want to put in the basement of the Pentagon where Antropic can't just switch off the ai.
They already have that. By definition. If Anthropic has done the work to be able to run on classified networks, then it's already running air-gapped and is not under Anthropic's control.
The thing is, just because you're in a SCIF doesn't (1) mean you can just break laws and (2) Anthropic don't have to support "off-label" applications.
So this is not about what they have and what it can do today - it's about strong-arming anthropic into supporting a bunch of new applications Anthropic don't want to support (and in turn, which Anthropic or it's engineers could then be held legally liable for when a problem happens).
Claude will answer all of those questions. The restriction Anthropic has is letting Claude pull the trigger and vibe-murder with no humans in the loop.
I reached to answer but idk what you mean by the second question. Long story short, Department of “War” wants Anthropic to say theres no restrictions on their use of Claude, Anthropic wants to say you can’t use Claude for domestic mass surveillance or automating killing people domestically or in foreign countries. Rest is just complication. And don’t peer too closely at the “Do”W”” wants Anthropic to say $X, the Team Red line (or, whatever’s left of them publicly after this last year) is basically “you can’t tell the gov’t what it can and can’t do, that’s it, it’s not that Do”W” will use it for that”
> Can someone in plain terms explain what this is really about?
This administration built almost entirely of dunces and conmen has convinced itself/been convinced that chatbots will help them in deciding where to send nukes, and/or they are invested in the incredibly over-leveraged companies engaged in the AI-boom and stand to profit directly by siphoning taxpayer dollars to said companies. My money is on the latter more than the former, but they're also incredibly stupid, so who's to say, maybe they actually think Claude can give strategic points.
The Republicans have abandoned any pretense of actual governance in favor of pulling the copper out of the White House walls to sell as they will have an extremely hard time winning any election ever again since after decades of crowing about the cabal of pedophiles that run the world, we now know not only how true that actually is, but that the vast majority are Conservatives and their billionaire buddies, and the entire foundation and financial backing of what's now called the alt-Right, with some liberals in there for flavor too of course.
If this shit was going down in France, the entire capital would have been burned to the ground twice over by now.
Or there won't be another election. They keep telling us there won't be another election. Why aren't we more alarmed by that? Why are we assuming they are lying about that?
Yes. All companies that deal with the government have agreed to let the government do whatever it wants within the bounds of whatever it is those companies do.
That's a shame. They might at least continue to work together to spy on foreigners. I don't understand the fuss anyway, what do claude models do that gpt and gemini can't?
As a foreigner, i see this as a great thing! I was about to cancel my Claude sub, but now i might hold on to it for a little and see how this plays out.
Future Trump rally: "And I hear Anthropic monkeyed with their dishonest chatbot Claude. They turned it Democrat! They trained it to say we lost the election against Sleepy Joe!"
Sure but I don't find them irreplaceable. Actually anthropic models have dropped out of my top ten usage this month. I only use opus occasionally for writing plans, its been pretty unreliable at executing.
A good reason to outlaw bribes is that politicians tend to be incredibly cheap and offer an extremely high ROI.
Albeit at the cost of a nice democracy.
You're forgetting that this is the same guy who managed to bankrupt a casino. He's not actually that good with money and until the latest bribe channels opened, eg Trump Coin and the Board of Peace, opened their finances may have been in a bit of a mess. Also I'd bet the ballroom donation was much larger, it's a massive blackhole of graft waiting to happen.
It's also not solely about money, you can get far just knowing how to chum it up with Trump when you get in the room with him. Look at the odd quasi-bromance between him and Mamdani who you'd expect to be enemy #1 but Mamdani knows how to schmooze the exact type of New York Guy Trump is.
For fascism, it's not always about getting something you think is a lot. It's about a power relationship. Trump has demonstrated that Nvidia will bow to his will.
It's also potentially an implementation of the foot-in-the-door technique (https://www.simplypsychology.org/compliance.html). It's a common manipulative strategy where you get someone to do a small favor for you which makes them much more likely to do a large favor for you later.
The branch of government tasked to execute the law has been ignoring laws. So we'll get a (from Trump's point of view) adversarial congress, so what, let's ignore them, what are they going to do about it?
Looking forward to a military platoon defying orders and seizing the president, hey, all countries suffer through coups, about time this young democracy go through one!
The terms of these markets do not account for a scenario, quite likely if authoritarian takeover does happen, where the House of Representatives is a rump organization which does not exercise effective power. There was a years-long period in Venezuela where the country's traditional legislature met and conducted business under the leadership of the opposition party, but actual legislative power was held jointly by the Supreme Court and a secondary legislature that Nicolas Maduro set up.
I feel this is a facile interpretation of the phrase.
A more steel-man interpretation of POSIWID--the way it's intended to be understood--would be: "What an established system accomplishes in the long term reveals something important of the the true preferences of the various interests that control it, which can easily diverge from the system's stated goals."
Great read. I've always noticed that the type of argument invoked is often less telling than when and in which context you invoke that argument.
You can make a lot of claims and they can match to reality a lot - normally people think of evaluating things in terms of a strict "does this fit or does this not", but it's often the meta-style (why do you keep bringing up that argument in that context?) that's important, even if it's not "logically bulletproof".
Wow that post is bad. The author clearly never actually attempted to understand what POSWID actually means and where it is coming from. Perhaps, instead of looking at Twitter, they should have opened Wikipedia. Or, better yet, Stafford Beers books (though admittedly, he was a pretty atrocious writer).
The follow-up is slightly better. But still not very convincing, IMO. They get far too stuck on a literal interpretation. Of something that self-describes as a heuristic.
The phrase does not make more sense even if we go all the way back to Beers. I certainly don't feel alone in not understanding how he went from his (fair) observation that "[There's] no point in claiming that the purpose of a system is to do what it constantly fails to do" to his more controversial conclusion: "The purpose of a system is what it does (aka POSIWID)".
Surely, there were many more sensible (but perhaps less quippy) stops between the two.
Trump wrote a long rant on Truth Social and ordered ALL federal agencies to stop using Anthropic. Not just the department of defense. This is straight up authoritarian.
Meanwhile, irrelevant "AI Czar" David Sacks, member of the PayPal mafia alongside known Epstein affiliates Elon Musk and Peter Thiel, is furiously retweeting all the posts from Trump, Hegseth, and other accounts. He is such a coward and anti American:
I don’t see a contradiction here. If control is out of the hands of decision makers, that’s a supply chain risk
. Were it not for that, the service is seen as critical to national security.
I dunno, safeguard seems like a weasel word here. It’s just reserving control to one party over another. It’s understandable why the DoD(W) wouldn’t like that.
That link is broken for me but I assume you meant to link to [0]. I think if there is a “safeguard” in a system, that definitely fits the bill of a supply chain risk. The only vague term here is “adversary”.
HN separates trailing dots from URLs, so that you can have working URLs at the end of a sentence. Hence you have to percent-encode trailing dots if they are a necessary part of the actual URL. (Same for some other punctuation characters, probably.)
This behavior is common for auto-hyperlinking of URLs in running text, so it’s bad practice to have such URLs.
Ugh, sorry for the broken link, I even pasted the same string into a new tab to make sure it worked because I thought the period at the end looked weird, and it was fine. Dunno how it got mangled.
[EDIT] Oh man, yours is like that too? WTF.
[EDIT2] If I follow your link, hit the 404 page, then add a period at the end of the URL, it does load. God that's strange.
"Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic."
This is authoritarian behavior. You're having trouble negotiating a contract, so instead of just canceling it - you basically ban all of F500 from doing business with that firm.
American people: latinamerican here. Maybe it's silly to root for a country in the world hegemony arena. I've usually been partial to the USA over China. Now I'm not rooting for your country anymore. As far as I'm concerned, I'd rather have China being the foremost power, at least they seem to be less keen on invading or heavily strong-arming latinamerica
That’s the crazy thing. This whole dispute was over Anthropic saying no to fully automated kill bots. They only required there be a human in the loop to press the button.
Anthropic didn't even say "no", it was more of a "not yet, let's work on this".
I really wonder what Palantir's role in all this is because domestic surveillance sounds exactly like Palantir and whatever happened during the Maduro raid led to Anthropic asking Palantir questions which the news reports is the snowball that escalated to this.
They also said no to fully automated AI domestic surveillance. I suppose non-US citizens like me are screwed but that's at least some small comfort for the natives. FVEY will just spy on each other and share but at least someone tried.
The ones that still assume big brother will be spying on and killing the people they hate. Trump openly campaigned on getting revenge on his enemies. I can only assume his supporters want this. The danger of course is if/when the leopards eat their faces
I guess the problem for Trump is if he orders the army to gun down protesters, there’s a good chance they will refuse to do it. While a bot can just be prompted to go ahead.
If we were able to give the Ukrainians fully automated kill bots, and those kill bots enabled Ukraine to swiftly expel the Russians from their territories, would that not be a good thing? Or would you rather the meat grinder continue to destroy Ukraine's young men to satisfy some moral purity threshold?
If we could give Taiwan killbots that would ensure China could never invade, or at least could never occupy Taiwan, would that be good or bad? I have a feeling I know what the Taiwanese would say.
While we're at it, should we also strip out all the machine learning/AI driven targeting systems from weapons? We might feel good about it, but I would bet my life savings that our future adversaries will not do the same.
Yes, I think Russia's invasion of Ukraine is quite clearly a binary Russia=bad, Ukraine=good. Same for the impending Chinese invasion of Taiwan. Perhaps you could explain the nuances under which Russia was the good guy? Better yet, maybe you could explain it to the Ukrainians who have been displaced, or the family members of those who have been killed, or the soldiers who have been permanently maimed?
Whether you or I like it or not, automatic kill bots will be a thing. It will only be a question of which countries have them and which do not.
And there is evidence automated killbots were already used in Gaza (not that that's a good thing).
Generally, in war, there are no rules, and someone is going to make automated killbots, and I expect one place to see them quite soon is in the Russia-Ukraine war. And yes, I'm hoping the good guys use them and win over the bad guys. And yes, there are good guys and bad guys in that conflict.
Ukrainian young (24 y.o.) man here. Living and working in police 30 kilometres away from the actual frontline.
No, thanks, we don't need those "fully automated kill bots". There's absolutely no guarantee that they wouldn't kill the operator (I mean, the one who directs them) or human ally.
We're pretty much fine with drone technology we have.
But for me personally, that's not the most important point. What is more important - and what almost no one in the Western countries seems to realise (no offence, but many of westerners seem to be kind of binary-minded: it's either 0xFFFFFF or 0x000000, no middle ground at all) - is that on the Russian side, soldiers are not "fully automated kill bots" either. Sure, there's a lot of... let's say - war criminals. Yes, for sure. But en masse they are the same young men that you can see on the Ukrainian side. Moreover, many people in Ukraine have relatives in Russia, and there already were the cases where two siblings were in different armies, literally fighting with each other. So in my opinion, "fully automated kill bots" are not an option here. At least unless you deploy them in Moscow and St. Peterburg to neutralize all of the Russian elites, military commandment and other decision-making persons of the current regime.
Fully automated kill bots are coming, whether any of us like it or not. The question is, which militaries will have them, and which militaries will be sitting ducks? China is pursuing autonomous weapons at full speed.
Personally, I think it'd be great to have the Anthropic people at the table in the creation of such horrors, if only to help curb the excesses and incompetencies of other potential offerings.
'yet'. Their reason for not allowing autonomous weapons usage was it isn't ready, not that they wouldn't do it on principle. Only the surveillance objection was on principle.
Sleep well in a box under the overpass maybe. If Amazon can’t serve Anthropics model until the courts get everything figured out it will be too late for them.
I’m just laughing at the possibility of it he US military being forced to use Chinese open source AI models because every US model provider refuses to work with them.
As written this would be the end of Anthropic. AWS, Microsoft et al are all suppliers of the DoW and as written they must immediate stop doing business with Anthropic. Will be interesting to see how this unfolds.
As a Canadian looking in, I see people talking about a 36% approval as low.
How is it that high!?
That means that more than 1-in-3 of your countrymen are ride-or-die, and it's just heartbreaking to see that we're going to have to launch that many people into the sun.
To counter point, do you think AI companies located on our adversaries turf will take the same stand? I agree its nightmarish to think of AI surveillance. But why is that being lumped in with weaponry? I see these as two separate issues.
Anthropic isn't even taking a particular hard stance. Their mass surveillance prohibition only applies to domestic spying, so they're a-OK with spying adversaries. If all of the AI companies all over the world took the same stance, it wouldn't improve the life of Americans one bit.
The only other thing that the foreign AI companies could do is say no to automated killing bots, which doesn't even seem like that good of an idea considering that your countrymen will most likely have to interact with these robots that can kill without any oversight.
He's more polarizing than usual maybe with stronger approve/unapprove ratings but his net popularity is in line with most 2nd term presidents at this stage.
Should military contractors put conditions on the use of their weapons?
Here's our tank, but you can't invade Iran with it?
We think your invasion of Venezuela is illegal, we're activating the kill switch on your jets.
That's a real dangerous proposition.
If the T&C is agreed to up front, why shouldn't they be able to? If their client or potential client doesn't like the T&C, they can find another vendor.
Ukrainians and Russians are experimenting with FPV drones using AI for target acquisition and homing. Not yet economically viable because it is cheaper to give your FPV fiber spool instead of Nvidia Jetson to bypass jamming.
When we have first politician blown to bits by autonomous AI FPV there will be sheer panic of every politician in the world to put the genie back into the bottle. It will be too late at that point.
Autonomous loitering munitions with 'AI' (image classification CNNs) are already in service and have been used - most demonstrably by the IDF.
Even during the Nagorno-Karabakh war, Azeri loitering munitions were able to suppress Armenian air defenses by hitting them when they rolled out of of concealment. I believe that killchain requires a level of autonomous functionality.
Azerbaijan was buying a lot of weapons from Israel prior to Nagorno Karabach war, so it is very likely that you have been talking about same weapon system in both cases.
However Russians and Ukrainians are using AI recognition in recon drones, but not yet in FPV. There is strong suspicion that long range one way attack drones are using AI during terminal guidance, but I did not see it confirmed by either side.
Decades of speculative science fiction, thought experiments, and discourse led to this. It’s gratifying to see that we’ve garnered enough concern, a major AI lab risking this to reign in the potential of runaway AI disasters. Hopefully we see other labs follow.
There is clearly a need to codify into all of these historical acts that they can't be invoked unless there is a declaration of war (or some other appropriate prerequisite).
This administration consistently exploits what were designed to be emergency powers because no such requirement exists. Leave no room for interpretation.
The current administration scoffs at laws. Nothing stopping them in that case from declaring war on Nauru and doing all the same. The solution is a sane, informed electorate, which is much more difficult in this age where a few disgustingly rich people have so much influence over news and media.
It seems like some comments here are from merged threads AND front-dated?
Makes for very confusing reading when comments from "1 hour ago" are actually on preceding events from earlier, before TFA news (announcement of designation).
mods: Especially in sensitive and rapidly developing situations like this, please don't mess with timestamps of comments. It's effectively revisionism.
This sounds like a message to would-be founders: don't base your company in the US. The strongest markets to do business are the ones with the most freedom from government meddling. In the US, big government is happy to use its power to crush private enterprise that it doesn't like.
Note that previously this label has been applied (nearly?) exclusively to non-US companies. US companies that don't do business with the DoD are not affected, and non-US companies that do business with the DoD are affected.
It's nice to see Anthropic sticking to their terms. I just have one question in all this. Why is Anthropic being singled out when it seems all the other big players are down to play with the DoD? Is this just a pissing match, or have the Anthropic models been proven the real winner for them?
It's same reason this administration recently tried to indict six Congresspersons for urging military members to resist "illegal orders." They want to demonize anyone who isn't blindly loyal to their side.
It may not be obvious. But this is actually a good thing when we looking back in a few years. I always feel weird that executive branch can just destroy private enterprise with "Supply-chain Risk" / "Terrorist List" without Due Process.
That's a good thing right? In a capitalist society, you cannot just burn $300B without consequences. Not to mention it is not just anyone's money. It is Saudi's.
> "Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic."
Does this mean Azure & AWS will have to stop offering Claude as a model?
I wonder, can't Amazon create a new legal entity to split AWS into "AWS-for-DoD" and "AWS-for-everyone-else"? So one can work with Anthropic and the other can't. Not sure how it works in the US.
A drunkard, ex-fox news host, wants mass surveillance and automated killing, what could go wrong?
I wish I thought enough Americans had the spine required to stand up to this, and I know for a fact that a lot do... the solution is literally written into your constitution.
Does Anthropic have standing to sue to Government for libel? I don’t think the Government is allowed to arbitrarily designate a company a supply chain risk without good cause.
Help me understand the line Anthropic is drawing in the sand?
Don't get me wrong i'm glad they are unwilling to do certain things...
but to me it also seems a little ironic that Anthropic literally is partnered with Palantir which already mass surveills the US. Claude was used in the operation in Venezuala.
Their line not to cross seems absurdly thin?
Or there is something mega scary thats already much worse they were asked to do which we dont know about I guess.
The whole reason this is happening is because Anthropic looked into how Claude was used in the Maduro op and found it to violate the negotiated terms of service.
Their hard lines are:
- no usage of AI to commit murder WITHOUT a human in the loop
What i'm saying is yes the UX is completely irrelevant and so is adding a human in the loop before murder.
A real stance would be not allowing your product to contribute to murder of any kind. I think focusing on the petty distinction they are making is missing the forest for the trees.
Thank you for the clarifications though it is helpful context.
Given that Anthropic is clearly risking their entire business just to stand up for what they believe is right, which appears to be what everyone here agrees with, is everyone who is supporting them here planning to also start using Anthropic and switch away from other vendors until they follow suit? Or are folks planning to just use whatever regardless?
Edit: I should perhaps clarify I'm more interested in paid users, rather than free. It's harder to tell if free users switching would help them or hurt them... curious if anyone has thoughts on that too.
> Our position has never wavered and will never waver: the Department of War must have full, unrestricted access to Anthropic’s models for every LAWFUL purpose in defense of the Republic.
Kesha tried to hug Jerry Seinfeld vibes.
> Anthropic delivered a master class in arrogance and betrayal as well as a textbook case of how not to do business with the United States Government or the Pentagon.
Strange way of saying "this vendor doesn't meet our software requirements".
> they have attempted to strong-arm the United States military into submission
Err... You approached them?
> a cowardly act of corporate virtue-signaling that places Silicon Valley ideology above American lives.
It's an orthogonal point, but "Silicon Valley ideology" has made up a significant portion of the USA's GDP for the last however many years.
> Their true objective is unmistakable: to seize veto power over the operational decisions of the United States military. That is unacceptable.
Again... You approached them?
> I am directing the Department of War to designate Anthropic a Supply-Chain Risk to National Security.
Like most companies in the world I imagine. They just haven't been approached yet.
> to allow for a seamless transition to a better and more patriotic service.
Internally re-framing all the recent "EU moving away from American tech!" articles as "EU builds more patriotic services!"
> This decision is final.
Nothing says "final" like a Tweet. The most uncontroversial and binding mechanism of all communication.
I imagine I'm not the only one to switch over to giving Claude my money today. I'm sure the "Other" comments for the cancellation were often as blunt as mine.
Q: "Is there anything we could do to change your mind?"
Don't worry, they will be seized by the government soon. Sounds crazy right. Not that far from the headline though, that would sound insane a mere 18 months ago.
The whole thing is fascinating. In my heart of heart, in principle, I want models to be essentially unrestricted, but I still find it somewhat problematic that government thinks it can say: you will make adjust your product to match our exact expectations even if you don't sign an updated contract with us. Odd stuff. I know they are trotting out War powers, but.. well.. we are not at war ( at least not yet or at least not yet officially declared.. ).
Working with the government is typically a huge pain in the ass unless you have a lot of friends on the inside. It's not hard to do the math when you you dealing with a government whose acting incredibly oppositional.
I had the co-founder of Levels and current head of the US Treasury Sam Corcos reach out to me a few weeks ago for a job. I was initially kind of excited because I had really wanted to work for the Treasury a couple years ago, so I took the phone call with him.
He called me and he seemed like a nice enough guy, but I realized that he's one of the DOGE/Elon acolytes and he started talking about how he's "fixing" the Treasury and that every engineer is apparently supposed to use Claude for everything.
It would have been a considerable pay downgrade which wouldn't necessarily be a dealbreaker but being managed by DOGE would be, but mostly relevant is that I found it kind of horrifying that we're basically trusting the entire world's bank to be "fixed" with Claude Code. It's one thing when your ad platform or something is broken, but if Claude fucks something up in the Treasury that could literally start a war. We're going to "fix" all the code with a bunch of mediocre code that literally no one on earth actually understands and that realistically no one is auditing [1].
If they're going to "fix" all the Treasury code with stuff generated by Claude, I'm not sure they will have a choice but to stick with it, because very it seems very likely to me that it will be incomprehensible to anything but Claude.
[1] Be honest, a lot of AI generated code is not actually being reviewed by humans; I suspect that a lot of the AI code that's being merged is still basically being rubber-stamped.
There's an awful lot of momentum with the USD being the world currency. Even if it eventually declines I think it might take decades, if the British pound is anything to go by.
I am fine with this. If you are a defense contractor, you are a defense contractor, and you follow the military needs that you government believes are necessary - or you stop being a defense contractor.
I wouldn't want a bullet manufacturer to hold back on my government based on their own internal sense of ethics (whether I agreed with it or not, it's not their place)
You're fine with a company being designated a supply chain risk, a designation heretofore used exclusively for foreign adversaries and usually a death knell for most companies, because the government wants to break a negotiated terms of service and contract that they already accepted?
Under normal circumstances this would end up in court, but when this administration ignores court orders it doesnt like Anthropic would effectively have no legal recourse.
If the government thinks the terms of Anthropic are unacceptable, they can just stop using them, right?
But why would you then retaliate and ban other companies from making business with Anthropic if they want to be a defense contractor?
How do these requirements make Anthropic a supply chain risk that makes them unusable for use by other companies?
> If the government thinks the terms of Anthropic are unacceptable, they can just stop using them, right
That is what they are doing.
> why would you then [....] ban other companies from making business with Anthropic if they want to be a defense contractor
Because, if it shops with Anthropic code, the DoD becomes subject to the restrictions when they receive the contractor's product. Anthropic's limitation is on the use, not on the product or distribution.
To stop using them requires making the suppliers still using them as well.
It's perfectly reasonable for the US government to end the contract if they no longer like the terms they agreed to (assuming the contract does in fact let them); it's not reasonable to destroy the counterparty to the contract in retaliation. The line "I am altering the deal; pray I don't alter it further" is literally spoken by Darth Vader, the most comic-book of comic-book villains.
> Anthropic’s stance is fundamentally incompatible with American principles.
I don't think that Secretary Hegseth is qualified to speak on American principles.
Cheating on multiple spouses[1], being an active alcoholic, and being accused of multiple sexual assaults and paying off the accusers[3] is fundamentally incompatible with being a Secretary of Defense and a good leader.
Also, this violates freedom of speech and will probably get shot down in the courts.
I got downvoted for this in the other thread, but this is basically an attempt at bankrupting Anthropic. No US company has ever been designated a supply chain risk, and the foreign companies that are on that list are now doing 0 business in the US. Very large portion of the US economy relies on some contracts with the US government, Anthropic cannot survive this if this holds.
I don't think it will hold, in the end this is mafia behavior, but if it does, we are yet again in uncharted waters.
I'd at least, you know, pretend we had a top-secret amazing model. By airing all of this publicly, they've basically admitted that Claude is the best there is.
Besides just being yet another example of the Trump admin abusing power and weaponizing legitimate laws in illegitimate ways to extract concessions, there is another reason this is dumb -- which is that Anthropic just has the best models!
As someone who wants America to win, ripping out Claude and putting in xAI is a terrible idea. Definitely setting us back a few months on capabilities
Bluster followed by a "we can't do it now but we will... soon". Whoever has the best model can do what they please you'll see. I work with these things daily as an engineer (been doing this shit for 25 years and wow it's like mana from heaven these days). Believe me no one is going to screw with themselves by not using the best one and right now Anthropic has it.
Anyone who does 996 is being exploited, unless they're the actual boss, in which case they're the ones doing the exploiting if they're pushing 996 on their employees.
This is why 996 bosses think AI can replace their employees, because they already see the employees as robots, not humans.
Its one thing to say "we cannot abide by these terms, so let's part ways", and its another entirely to respond this drastically. The Trump administration will look back on this decision as the most consequential in their efforts to win the 2026 midterms and Republican efforts in 2028. This is a $400B+ American company that has significant partial ownership from Amazon, Google, and other private equity sources; they just made serious enemies in SV, many of whom supported Trump in his 2024 election victory.
This is a pimple on the arse of said consequence. It's one tiny thing in a chain of many bigger things.
It's magnified because it's right now, but this won't affect midterm results barely a whisker compared to many other daily headlines.
There are no serious enemies to this administration in SV and I can't see this changing that. SV has bent the knee exactly like Anthropic didn't. They're not going to stand up because of this, they've proven they don't have those muscles.
OTOH it could amplify their base: “Big Tech refusing to work with us on National Security matters!” The base will never hear what/where the red line was drawn, just that Some Company in California (liberal/bad) is being Woke and Political.
I can't seem to find what being designated a "Supply-Chain Risk to National Security" implies from a legal standpoint. From what I can find, it doesn't seem to be a formal legal status. Curious if anyone knows more.
Basically, if you are a federal contractor, the designation means the DoD can force you to certify that Anthropic tech is not used in the fulfillment of your government work. Because it's just a DoD designation, and an executive order and not added to the NDAA, you can still use Claude for non-government (federal) touching work.
So using Claude Code to write software for the DoD is now a no go, you'd be in breach of procurement directives now.
If they go as far as to convince congress to add Anthropic to the NDAA, that would be a nationwide ban like Huawei making it illegal for any federal contractor to use the tech anywhere in their business.
But for now, even fed contractors can still use Claude in their business, just not directly for government work.
That doesn’t seem to match up with the original tweet though - it sounds a heck of a lot stronger:
> Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic
Emphasis mine.
And I’m looking at news organizations that presumably have staffs of legal analysts pouring over this stuff, and they also seem to be saying that it can’t be any commercial activity:
> The label means that no contractor or supplier that works with the military can do business with Anthropic.
I don't think it's ever been about strong or weak, or at least I don't think that's where the differentiation is. You always want 'strong' government, committed to the things it says it's committed to.
It's more been about the size of the government; that it should do a minimal amount of control (and do it well), but leave a lot of things for "the market to decide".
Having said all that, I think this issue is just tangential to any big/small government ideology. This is a hissy fit about a defence contractor sticking to their agreement where the DoD want to change the agreement in a way that goes against the contractors Mission Statement and/or the US Constitution itself.
The old ideology of the
Republicans doesn't mean anything here. This administration is purely about 'give me what I want, now!'.
And it's whims change with the breeze. Do not look for consistency here.
How 'bout that government meddling in the free market, eh?
Every conservative needs to do some very deep, very serious soul-searching. As for me, as a hyper-progressive, I'm drawing up proposals for nationalizing real estate developers in order to force them to build new houses to sell below cost.
No, stop, I understand the politics here, but I’m asking about the technical fundamentals.
LLMs produce output of unknowable and unpredictable accuracy, and as far as we know, this is a mathematically unsolvable problem. This shit should not be within 1000 miles of a weapons system. Why are we even talking about this?
>Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic.
Nevermind Claude, does that mean Anthropic's offices can't use a power company if that same company happens to supply electricity to a US military base? What about the water, garbage disposal, janitorial services? Fedex? Credit card payments? Insurance companies? Law firms? All the normal boring stuff Anthropic needs that any other business needs.
This is a corporate death penalty. Or corporate internal exile or something, I don't know of a good analogy.
A level up, this is only the beginning of the political headwinds for AI. There will be a lot more, especially if constituencies begin to get displaced. I don’t think “job loss” will really occur, at least not in a dramatic way overnight. But I do believe there will be both aggressive regulation and very aggressive taxation of this technology in the near/mid-term.
Good. At least now I don't have to worry that my vibe-coded, unreviewed checkout button is accidentally going to hallucinate the command that blows up a kindergarten in Yemen.
We can actually get a glimpse of how AI might wipe out humanity here.
Model collapse making models identify everyone as a potential threat who needs to be eliminated.
Companies should have a right to refuse such requests on moral grounds though.
This stance is vindictive. Just don't use Claude in the military. Extending it to all government agencies is not right. They do great work. Can't deny that.
The US is such a shit show. Personally I hope this doesn't affect Anthropic's growth and development because I quite enjoy using their products and see them evolve.
This is only the first year of this fascist government, and I believe the first powerful company that is taking a stance? Meta, Apple, etc. have all bent the knee right?
Department of War is a teenage boy's idea of "manly" and "cool". Same with X. These juvenile idiocrats will be laughed at by children in the future studying history. "Seriously? How dumb were these people in the 21st century."
Already there 'February 23, 2026: The Pentagon confirmed a new agreement allowing Grok use in classified systems. Defense Secretary Pete Hegseth announced it would go live soon on unclassified and classified networks, alongside other models, as part of feeding military data into AI.'
This will mean Grok becomes the defacto US Gov AI provider.
Why are so many adopting this name for what is by law, by the American people, called the Department of Defense? The name change pertains directly to the Anthropic issue, which is the function of the government and department, the power of the American people to govern themselves, and the role of the president relative to the soveriegn American people.
Well put and it bothers me too. It seems to be another case of Orwellian manipulation, i.e. an expression of power through language, functioning as a litmus test of the speaker's loyalty. Serious publications are not going along with it. More craven or (here) thoughtless ones are falling in line.
I mean the original name switch was much more "Orwellian manipulation" if anything changing it back to war is undoing the bullshit implications that everything it does is defense.
Surely the purpose of the organization is to defend the country? War seems more like the failure mode. The point here is that it was established by a law of Congress and so has an official name that should be respected until another law changes it.
OpenAI came out just last night or today claiming they would hold the same line as Anthropic. Makes me think both sides knew Elon had already won the contract.
And here’s the irony: Musk, who claimed only he is virtuous enough to defend us from AI, who insisted he always wanted model labs to be non profit and research focused, will now bring his for profit commercial entity into service to aid in mass domestic censorship and fully autonomous weapons of war.
In fact it won’t surprise me further if NVIDIA is strong armed into providing preference to xAI, in the interest of security, or if the government directly funds capital investments.
Anthropic saves some dignify and they’re the losers today, but we are the losers tomorrow.
I can't wait to read the transcript of the AUSA in front of a federal judge trying to explain threatening to declare a company a supply chain risk if the company doesn't supply things to the government.
> 2. No fully autonomous weapons (kill decisions without a human in the loop)
Surveillance takes place with or without Anthropic, so depriving DoW of Anthropic models doesn't accomplish much (although it does annoy Hegseth).
The models currently used in kill decisions are probably primitive image recognition (using neural nets). Consider a drone circling an area distinguishing civilians from soldiers (by looking for presence of rifles/rpgs).
New AI models can improve identification, thus reducing false positives and increasing the number of actual adversaries targeted. Even though it sounds bad, it could have good outcomes.
But compared to what - if Anthropic's models aren't perfect but still better than existing (old school) models, it's understandable DoW still wants to use them (since they're potentially the best available, despite imperfections). I think Hegseth is saying to Anthropic: "that's our call, not yours".
But surely if Anthropic thinks there's a risk that their models might make bad decisions, and the resulting civilian or etc deaths are blamed on them, it's their right to refuse to sell it for that purpose? That's why they had those restrictions in the contract to begin with. How can they be forced to provide something?
I agree they can't be forced to provide something. I just see DoW's reasoning, and I can't fault it.
Anthropic are taking a moral position which is admirable, but in this case it could actually make people's lives worse (if we assume more false positives and fewer true positives, which is probably a fair assumption given how much better 'modern' AI is compared to the neural net image recognition of just a few years ago).
Batshit situation, respectable position from Dario throughout.
But there's some irony in this happening to Anthropic after all the constant hawkish fearmongering about the evil Chinese (and open source AI sentiment too).
Hegseth's had a busy week: trying to kill Anthropic, attending the State of the Union, fighting Scouting America, and his regularly scheduled efforts to shame fatties & trans kids... Unlike so many in the orange one's inner circle who are just incompetent (say, Kash Patel for one), this dude is both incompentent a very bad, bad person.
I think odds are high a lot of these posts are by staffers. The posting volume is bananas, even granting that he spends a lot more time personally online and watching cable news et c. than any prior president, I don’t think there’s any way they’re all by him.
I do think a lot of the more hot-take type posts (often in response to stuff he’s watching on tv) are either actually him, or he’s dictating to an aide. These larger policy-type ones that he treats as quasi-executive-orders, I think are likely drafted by one or more of his cabinet-level folks, or others roughly as high up. That’s just my speculation based on reading the “tea leaves”, though.
As for official word, it waffles between “all of it’s him” and “oh not that one though, that racist video repost was a staffer who made a mistake”, so that’s little help in sussing out the truth (but I am rather certain they’re not all directly written and posted by him)
This is going to have two unintended consequences.
One, it’s going to fuck with the AI fundraising market. That includes for IPO. If Trump can do this to Anthropic, a Dem President will do it to xAI. We have no idea where the contagion stops.
Two, Anthropic will win in the long run. In corporate America. Overseas. And with consumers. And, I suspect, with investors.
A lot of corporate America contracts for the military in some capacity (it's a giant piggy bank and if you jump through a few hoops you get to siphon money out of it, so of course they do) and assuming this Tweet is accurate (Jesus, what a world) this will also affect them.
IDK maybe they have corporate structures that avoid letting this kind of thing mess too badly with the parts of their company that don't have contact with the government, or maybe it'll only apply to specifically the work they do for the government, but otherwise I expect it'll be devastating for Anthropic's B2B effort.
AI proponents have been very vocal about AI safety being meaningless. But nobody could have expected that the end of the world would have come because Trump puts Grok in charge of the US nuclear arsenal. We truly live in the dumbest timeline.
"THE UNITED STATES OF AMERICA WILL NEVER ALLOW A RADICAL LEFT, WOKE COMPANY TO DICTATE HOW OUR GREAT MILITARY FIGHTS AND WINS WARS! That decision belongs to YOUR COMMANDER-IN-CHIEF, and the tremendous leaders I appoint to run our Military.
The Leftwing nut jobs at Anthropic have made a DISASTROUS MISTAKE..." - President Donald J. Trump
No surprise here. All government actions are now in the Trump mafia boss style.
“You won’t let us use your product unrestricted for military applications? Fuck you, we’re going to stop using it for anything at all across the entire federal government, even if not remotely related to military.”
Kudos to anthropic for standing up for their principles. Let's remember all the silicon valley leaders who embraced fascism without even needing to be pressured. We need more billionaires with backbones.
As best I can tell, his hard-drinking era ended many years before he entered the cabinet. But this does feel like a pretty impulsive decision, and there's some ambiguity over whether this statement was approved by the WH, or whether this was just the SECDEF taking it to the next level to look super loyal and badass. This ambiguity gives the WH room to walk it back in the coming weeks, depending on how things evolve.
I can honestly understand both positions. The U.S. military must be able to use technology as it sees fit; it cannot allow private companies to control the use of military equipment. Anthropic must prevent a future where AIs make autonomous life and death decisions without humans in the loop. Living in that future is completely untenable.
What I don’t understand is why the two parties couldn’t reach agreement. Surely autonomous murderous robots is something U.S. government has interest in preventing.
> it cannot allow private companies to control the use of military equipment.
The big difference here is that Claude is not military equipment. It's a public, general purpose model. The terms of use/service were part of the contract with the DoD. The DoD is trying to forcibly alter the deal, and Anthropic is 100% in the clear to say "no, a contract is a contract, suck it up buttercup."
We aren't talking about Lockheed here making an F-35 and then telling the DoD "oh, but you can't use our very obvious weapon to kill people."
> Surely autonomous murderous robots is something U.S. government has interest in preventing
After this fiasco, obviously not. It's quite clear the DoD most definitely wants autonomous murder robots, and also wants mass domestic surveillance.
Because the current government wants unquestioning obedience, not a discussion (assuming they were capable of that level of nuanced thought in the first place). The position of this government is "just do what I say or I will hit you with the first stick that comes to hand".
If the government doesn't want to sign a deal on Anthropic's terms, they can just not sign the deal. Abusing their powers to try to kill Anthropic's ability to do business with other companies is 10000% bullshit.
> What I don’t understand is why the two parties couldn’t reach agreement. Surely autonomous murderous robots is something U.S. government has interest in preventing.
Consider the government. It’s Hegseth making this decision, and he considers the US military’s adherence to law to be a risk to his plans.
I can see both sides as pertains to Trump's initial decision to stop working with Claude, but now, this over-the-top "supply chain risk" designation from Hegseth is something else. It's hard to square it with any real principle that I've seen the admin articulate.
> What I don’t understand is why the two parties couldn’t reach agreement.
Someday we'll have to elect a POTUS who is known for his negotiation and dealmaking skills.
The disconnect here for me is, I assume the DoW and Anthropic signed a contract at some point and that contract most likely stipulated that these are the things they can do and these are the things they can't do.
I would assume the original terms the DoW is now railing against were in those original contracts that they signed. In that case it looks like the DoW is acting in bad faith here, they signed the original contact and agreed to those terms, then they went back and said no, you need to remove those safeguards to which Anthropic is (rightly so) saying no.
Am I missing something here?
EDIT: Re-reading Dario's post[1] from this morning I'm not missing anything. Those use cases were never part of the original contacts:
> Two such use cases have never been included in our contracts with the Department of War
So yeah this seems pretty cut and dry. Dow signed a contract with Anthropic and agreed to those terms. Then they decided to go back and renege on those original terms to which Anthropic said no. Then they promptly threw a temper tantrum on social media and designated them as a supply chain risk as retaliation.
My final opinion on this is Dario and Anthropic is in the right and the DoW is acting in bad faith by trying to alter the terms of their original contracts. And this doesn't even take into consideration the moral and ethical implications.
[1]: https://www.anthropic.com/news/statement-department-of-war
The writeup here[1] was pretty clear to me.
> *Isn’t it unreasonable for Anthropic to suddenly set terms in their contract?* The terms were in the original contract, which the Pentagon agreed to. It’s the Pentagon who’s trying to break the original contract and unilaterally change the terms, not Anthropic.
> *Doesn’t the Pentagon have a right to sign or not sign any contract they choose?* Yes. Anthropic is the one saying that the Pentagon shouldn’t work with them if it doesn’t want to. The Pentagon is the one trying to force Anthropic to sign the new contract.
[1]: https://www.astralcodexten.com/p/the-pentagon-threatens-anth...
I just wish there was a stronger source on this. I am inclined to agree you and the source you cited, but unfortunately
> [1] This story requires some reading between the lines - the exact text of the contract isn’t available - but something like it is suggested by the way both sides have been presenting the negotiations.
I deal with far too many people who won't believe me without 10 bullet-proof sources but get very angry with me if I won't take their word without a source :(
I think a big question mark here, is whether anything said on Anthropic's side if in the framing of "We have a thing going on that we are trying to communicate around where a canary notice if it existed would no longer be updated"
That's a fair point, but I think Dario's quote in GP corroborates ACX's story quite well:
> "Two such use cases have never been included in our contracts with the Department of War..."
Also, Trump's own words complaining about being forced to stick to Anthropic's terms of service:
> The Leftwing nut jobs at Anthropic have made a DISASTROUS MISTAKE trying to STRONG-ARM the Department of War, and force them to obey their Terms of Service instead of our Constitution.
This administration needs the benefit of the doubt always. This administration deserves the benefit of the doubt never.
Those people are dealing with you in bad faith, and you need to cut them off before they try to overthrow your government again.
It isn't about commercial agreements, it's about patriotism. The national industry is supposed to submit to the military's wishes to the extent that they get compensated. Here it's a question or virtue.
The Pentagon feels it isn't Anthropic to set boundaries as to how their tech is used (for defense) since it can't force its will, then it bans doing business with them.
I'm guessing you're being down voted because people don't know if you think that's a good thing or not. I do not think it's a good thing. Do you?
I absolutely do not think that's a good thing. Was stating some sad facts.
No one cares if the Pentagon refuses to do business with Anthropic. But Hegseth has declared that effective immediately, no one else working with the DoD can either--which includes the companies hosting Anthropics models (Amazon, Microsoft, and Alphabet).
So it's six months to phase out use of Anthropic at the DoD, but the people hosting the models have to stop "immediately".
Which miiight impact the amount of inference the DoD would be able to get done in those six months.
>The national industry is supposed to submit to the military's wishes to the extent that they get compensated.
According to whom?
Regardless of the original contract, it's entirely appropriate for a vendor to tell the customer how to use any materials.
Imagine a _leaded_ pipe supplier not being allowed to tell the department of war they shouldn't use leaded pipes for drinking water! It's the job of the vendor to tell the customer appropriate usage.
This is quite literally the norm for things with known dangerous use cases.
Go look at the package on a kitchen knife and it says not to be used as a weapon
Playing devil's advocate: if I did in fact grab one of my kitchen knives to defend myself against a violent intruder into my kitchen, I wouldn't expect to be banned from buying kitchen knives.
I'm not sure this is still a useful analogy, though...
If I shoot someone, something that is explicitly warned against in firearm safety materials that come with every purchase of a new firearm, I am no longer allowed to purchase any more firearms.
That's for a different reason though--you broke the law.
And if you grabbed the knife and went on a violent spree, I'd absolutely expect the knife manufacturer to refuse to sell to you anymore.
The knife manufacturer isn't obligated to sell to you in either case, I'd expect them not to cut ties with you in the self defence scenario. But it is their choice.
The specific shape of a kitchen knife would make it a particularly poor fighting knife, and knives in general are bad for self defense, due to the potential for it to be turned against the user. So, there is a good argument that such a suggestion is really in the user's best interest rather than a cynical play for the manufacturer to limit liability.
These knife and lead analogies don't map well to the reality of AI. Note: just talking about the analogy itself not the point you are making.
Not in software though. Clear precedent has been established via EULAs. Software companies set the rules and if users don't like, they can piss off. I don't see why it would be any different for the government.
I'm not a fan of EULAs, I think if you acquire some software anonymously and run it on your own systems you should be able to do whatever you want. however if you want software hosted on someone else's machines, or want to enter into a contractual relationship with them then government or not you should not have the right to compel work from them.
Depending on the country, their legal value is limited: https://en.wikipedia.org/wiki/End-user_license_agreement#Enf...
A lot of things are different when it comes to national security, and military.
Congress could come up with an act it it's for national interest.
The military isn't the typical End User.
The government is armed and can exempt itself from prosecution either by judicial means and/or by naked force. So it isn’t just a cut and dry licensing problem.
Because it's the government? Companies need to follow the rules the government sets, if they like it or not
The government cannot set arbitrary rules, it has to follow the law. (And, at least with a functioning separation of powers, it cannot change the law arbitrarily.)
Um. No, that's not how it works...
They also have other vendors.
Claude Opus is just remarkably good at analysis IMO, much better than any competitor I’ve tried. It was remarkably good and complete at helping me with some health issues I’ve had in the past few months. If you were to turn that kind of analytical power in a way to observe the behaviour of American citizens and to change it perhaps, to make them vote a certain way. Or something like - finding terrorists, finding patterns that help you identify undocumented people.
I don't believe they can change the name to Department of War without an actor Congress. It remains the DoD.
Yes, it's officially still the Department of Defense.
If this were a news outline writing "Department of War" I would be concerned. But in the case of the Anthropic CEO's blog post, I can understand why they are picking their fights.
I first read about DoW on a post by Anthropic and thought it was some kind of jab to the government.
Well I think we have an actor congress
They can, however, rename their Twitter/X accounts and vacate the @SecDef handle, which seems to be up for grabs now, if anyone wants to do the funniest thing...
Of all the silly things that Trump did, I think this one is the most reasonable. This has always been a department of war. Calling it defense was propaganda.
Calling it Department of The Armed Forces or Department of Military would be neutral. Putting War in the name is as propaganda-like as Defense.
Gulf of America and department of war are nothing but propaganda and dick measuring. Prove me wrong please.
It's a silly shibboleth, but I automatically ignore anyone who calls it the Department of War or Gulf of America. Hasn't steered me wrong yet. They're telling me they're the kind of people who only care about defending fascism.
I call it department of war, because I think it is a great self-own on their part to do such a rename.
There will be no fighting in the war room!
I think it's worth giving people a tiny bit of grace on this. I've surprised people by explaining that the "Department of War" is just fascist fanfic and that the legal name has not changed.
It's a testament to the broken information ecosystem we're in that many people genuinely don't know this. Most will correct themselves when told. I agree with you that those who don't are not worth engaging.
Google Maps calls it Gulf of America, pretty difficult to ignore Google.
Only in America, in the rest of the world Google calls it "Gulf of Mexico (Gulf of America)".
Don't deadname the Gulf!
https://s3.gtw.lt/lUew91t6v5AO2u6mAPCXAFME.png
Depends where you at
I ignore Google quite easily. Besides, as soon as Trump is out they will change the name back.
Because Google are bootlickers.
They literally complied with this request immediately and without question.
It's almost like the democratically elected government gets to decide the name, not Google!
I would not defend all of Google's decisions in the Trump era, but complying immediately with politicized name changes has always been the status quo. Even in healthy democracies, the precise names of geographic features can be extremely controversial, and no sane company wants to get in a debate with the Japanese government about the real names of various islands.
the entire administration negotiates in bad faith. literally every agreement they sign whether it's international trade or corporate contracts is up to the whim of a toddler with twitter
You pretty much nailed it. I can't even get outraged at any given instance now that the trendline is so staggeringly clear.
I can't see anyway this ends well for the US. I say this as both an American and a military veteran.
It's the Department of Defense.
[1] "only an act of Congress can formally change the name of a federal department."
With this administration, after all their proven lies, when in doubt, assume bad faith on their part. Assuming good faith at this point is Lucy and Charlie Brown and the football, but now the football is fascism (i.e., state control of corporations, e.g., what Trump administration is doing here).
Trump has historically stiffed his contractors. Why do you think his administration would be any different with adhering to a contract?
I was pondering the same thing and to me the answer is a contractor sold something to the DoD and Anthropic pulled the rug out from under that contractor and the DoD isn't happy about losing that.
My speculation is the "business records" domestic surveillance loophole Bush expanded (and that Palantir is build to service). That's usually how the government double-speaks its very real domestic surveillance programs. "It's technically not the government spying on you, it's private companies!" It's also why Hegseth can claim Anthropic is lying. It's not about direct government contracts. It's about contractors and the business records funnel.
Yes, I assumed a mass surveillance Palantir program also. Interesting take on how it allows them to claim “we are not doing this” while asking Anthropic to do it.
Of course they can just say - we aren’t, Palantir is.
Network recon [0]
[0] - https://news.ycombinator.com/item?id=47180540
Wow, and the only restrictions Anthropic asked for are (1) no mass domestic surveillance and (2) require human-in-the-loop for killing [1]. Those seem exceptionally reasonable, and even rather weak, lol :|
[1] https://www.anthropic.com/news/statement-department-of-war
Did the DoW ask for these things?
This whole thing seems like people talking past each other, and that there’s something being left unsaid.
Anthropic doesn’t make a product that would assist with kill drones, and they don’t have the right to deny subpoenas.
Anthropic specifically called out systems "that take humans out of the loop entirely and automate selecting and engaging targets".
I take that to mean they don't want the military using Claude to decide who to kill. As a hyperbolic yet frankly realistic example, they don't want Claude to make a mistake and direct the military to kill innocent children accidentally identified as narco-terrorists.
At least, that's the most charitable interpretation of everything going on. I suspect they are also worried that the sitting administration wants to use AI to help them execute a full autocratic takeover of the United States, so they're attempting to kill one of the world's most innovative companies to set an example and pressure other AI labs into letting their technology be used for such purposes.
Right. Did the DoW ask for that? Or does Anthropic make a product that does that?
Their intention is to turn it against the American people. Hegseth literally wrote a book about eliminating democrats from the US, and this surprises people.
So they are such a risk to national security that no contractor that works with the federal government may use them, but they're going to keep using them for six more months? So I guess our national security is significantly at risk for the next six months?
It's a waste of your effort to apply rational argument to the actions of a group that are in it for a shakedown.
Simple rational argument:
SCOTUS says POTUS is above the law, so POTUS has collected $4B in bribe / protection money since taking office 13 months ago. Anthropic has lots of money at the moment. Why should they be allow to keep it?
Since they didn't pay off the president (enough?), his goons are going to screw with their revenue and run a PR smear campaign.
Once you realize it only has to do with Trump's personal finances, and nothing to do with national security or the rule of law, then all the administration's actions make perfect rational sense.
Open question: How much should a congress-critter charge Trump for a favorable vote? (The check should come with a presidential pardon in the envelope, of course...)
It’s the mob. This is nothing more than, “Nice AI ya got here. Be a shame if sometin’ wuz to happen to it.”
Except that it’s sovereign.
Sovereign like King George III?
Keep in mind that Anthropic “is the only A.I. company currently operating on the Pentagon’s classified systems” [1].
[1] https://www.nytimes.com/2026/02/27/technology/defense-depart...
They are the same amount of ‘risk’ to national security that the various ‘emergencies’ the executive branch has used as legal excuses to do otherwise illegal things are emergencies.
Congress is negligent in not reigning this kind of thing in. We’re rapidly falling down so many slippery semantic slopes.
I'm def adding "slippery semantic slopes" to my vocab.
the administration which declares ad-hoc emergencies is behaving as predicted
Can't just unplug the thing and use something else.
Obviously the DoD would not want limited use. Strange they don't make their own given their specific needs.
Isn't this our governments classic negotiation strategy? Go to the extreme, and meet somewhere well on their side of the middle.
The Trump administration tends to use this playbook.
Putting aside my take, I’m trying to objectively make sure I’m grounded on what is likely to happen next, without confusing “what is” with “what is ok”.
Don't make the mistake of thinking their words have meaning. They see a way to punish the company, they take it. Same thing with declaring a national emergency to impose tariffs. There's no supply chain risk, no national emergency, but that doesn't stop them.
Dont forget Nvidia technology was condsidered too sensitive to be exported to China....until the Trump administration decided they could export it if they paid a 10% export tax.
We've moved beyond telling people not to forget and have entered "expect nothing less" territory
Aren't export taxes against the US constitution?
Yes ... but what's your point? /s
> So I guess our national security is significantly at risk for the next six months?
That does seem to be what Hegseth is arguing, yes; and that is presumably his justification for doing something drastic here. Although I assume he is lying or wrong.
And as a cynic, let me just add that the image of someone going to the political overseers of the US military with arguments about being "effective" or "altruistic" is just hilarious given their history over the last ~40 years.
There has been a terrifying decline in quality and an increase in corruption in Trump’s second administration.
Re: the hilarity part, I’m conflicted: in general, a good sense of humor is useful, but in present circumstances a stoic seriousness seems warranted.
I admire Anthropic for sticking to their principles, even if it affects the bottom line. That’s the kind of company you want to work for.
Companies change (remember "don't be evil"?) but yeah for the Anthropic of today, respect.
It's also a very clear differentiator for them relative to Google, Facebook, and OpenAI, all of whom are clearly varying degrees of willing to sell themselves out for evil purposes.
It will also cost openai dearly if they don't communicate clearly, because I for one will internally push to switch from openai (we are on azure actually) to anthropic. Besides that my private account also.
This will not cost OpenAI anything.
You can deploy Opus and Sonnet on Azure.
Is making effective weapons evil?
Given the history of US military adventurism and that we’re about to start another completely unjustified war of aggression against Iran, yes. Absolutely yes.
That’s a simplistic framing (obviously)
What does effective weapons mean in this particular instance?
Yeah
Yes?
"You need me on that wall!"
The team that handles their PR has done an amazing job in the last 9 months
Hint: It's much easier to have good PR by being actually good. Though it does make people like this do the whole implication thing.
I saw this the other day:
> Costco is a really popular subject for business-success case studies but I feel like business guys kinda lose interest when the upshot of the study is like "just operate with scrupulous integrity in all facets and levels of your business for four decades" and not some easy-to-fix gimmick
https://bsky.app/profile/mtsw.bsky.social/post/3lnbrfrvmss26
I don't know, staff at my two Costcos feel much more disinterested and rude then I remember a decade ago. It used to feel fun but now it's miserable.
At peak times they run out of carts and tell the customers to go hunting in the lot for them, door greeters shouting at members across the floor, checkout queues stretch the length of the warehouse, they start half blocking the gas station entrance 30mins before close so trucks can't get in, so maybe they're turning those profit screws.
>It used to feel fun but now it's miserable.
It's not their job to entertain you.
Ah, right, by being actually good, as in - being okay with mass surveillance as long as it isn't being done in the US, being okay with Claude assisting in killing people as long as it isn't fully autonomous, and being actively hostile to open-weight LLMs and open research on LLMs? This kind of "good"?
No, OP is right, their PR department is doing a great job.
Correct. Protect our citizens' rights, as we are the ones under the jurisdiction of our government. Yes, design competitive weapons systems that can stand up to the threats that adversary powers are creating, but do so while maintaining human control.
That kind of good.
It’s nice that Americans are being so open about how they feel about other countries these days.
This whole saga is extremely depressing and dystopic.
Anthropic is holding firm on incredibly weak red lines. No mass surveillance for Americans, ok for everyone else, and ok to automatic war machines, just not fully unmanned until they can guarantee a certain quality.
This should be a laughably spineless position. But under this administration it is taken as an affront to the president and results in the government lashing out.
If you're a billionaire there's no risk to "sticking to principles", so there's nothing to admire. Also that's not what they're doing. These are calculated moves in a negotiation and the trump regime only has 3 years left. Even a CEO can think 4 years ahead.
It's probably in Anthropic's interest to throw grok to these clowns and watch them fail to build anything with it for 3 years.
i disagree. 3 years is an insanely long time in the AI space. The entire industry pretty much didn't even exist three years ago! Or at least not within 4 orders of magnitude.
Also, every other company has bent the knee and kissed the ring. And the trump admin will absolutely do everything they can to not appear weak and harm Anthropic. If it was so easy to act principled, don't you think other companies would've refused too? Eg Apple
And there is real harm here. You're reading about it - they get labeled a supply chain risk. This is negative and very tangible
Considering how many bootlicking billionaires I see these days, it is still a bit surprising.
Exactly.
> 83 people in total killed in US attack to abduct President Nicolas Maduro
Blood is on their hands already
Good. I'd rather not have my favorite AI from a company working on AGI to have murder and spying in it's DNA.
In fact, as a patriotic American veteran, I'd be ok with Anthropic moving to Europe. It might be better for Claude and AGI, which are overriding issues for me.
Rutger Bregman @rcbregman
This is a huge opportunity for Europe. Welcome Anthropic with open arms. Roll out the red carpet. Visa for all employees.
Europe already controls the AI hardware bottleneck through ASML. Add the world's leading AI safety lab and you have the foundations of an AI superpower.
https://x.com/rcbregman/status/2027335479582925287
> Good. I'd rather not have my favorite AI from a company working on AGI to have murder and spying in it's DNA.
Anthropic made it quite clear they are cool with spying in general, just not domestic spying on Americans, and the "no killbots" pledge was asterisked with "because we don't believe the technology is reliable enough for those stakes yet". The implication being that they absolutely would do killbots once they think they can nail the execution (pun intended).
I suppose you could say they're taking the high road relative to their peers, but that's an extremely low bar.
I wouldn't say it's clear. People keep pointing to the wording used in the statement to say it, but I wonder if it has to do with constitutionally; domestic surveillance of people in the US without a warrant is against the constitution, and surveillance of non-citizens outside the U.S is not. Can they even be compelled by the executive branch to do an action that may be unconstitutional?
Do all of the employees want to move to Europe suddenly? Unless it’s the UK or Ireland, do they speak the local language? If it is the UK or Ireland, do they prefer the weather in California? Do they have children in school or in college locally? Do they have family they’d rather not move 9 time zones away from? Elderly parents they’re taking care of?
They only have to move their headquarters no? Reincorporate in France. Hire Yann LeCun (I like LeCun)
I'm pretty vocal about our collective authority to work against the Trump administration, and even I would be hesitant to work as a US employee of a company that fled the country after a dispute with the US military. Seems like an extreme threat to my personal safety for little resistance benefit.
Canada is another option. Canada has significant AI research institutes going back decades ( https://en.wikipedia.org/wiki/Mila_(research_institute) ) that have produced much of the foundational research that backs today's AI models.
For Americans and international researchers it's easy to get visas there quickly. It's not far at all for Americans to relocate to or visit. Electricity is cheap and clean. Canada has the most college educated adults per capita. The country's commitment to liberalism, and free markets, is also seeming more steadfast than the US at this point in time.
Canada faces obstacles with its much smaller VC ecosystem, its smaller domestic market, and the threat of US economic aggression. Canada's recent trade deals are likely to help there.
I say this all as an American who is loyal to American values first and foremost. If the US wants to move away from its core values I hope other countries, like Canada or the EU, can carry on as successful examples for the US to eventually return to.
If Anthropic moving to Europe was better for Claude, why has Europe not produced Claude?
I have my doubts about Anthropic wanting to pick up and move the entire company to Europe even if Ursula von der Leyen personally signed their visas. Maybe only if the government tried to nationalise their proprietary models.
doesn't the Defense Production Act essentially do that?
Why wouldn’t the government just arrest their board and execs on charges of treason or something? At this point they could probably publicly hang them all and a plurality of Americans would cheer it. I don’t know if you appreciate how disliked tech is by the left and right alike.
The left would never support that lawlessness: opposition to AI is based on things like ethics, environmental impact, etc. which are predicated on concepts like the rule of law. People are calling for regulation or UBI, mor killings.
The right has far more talk of violence, true, but a lot of that is targeted rhetoric to keep voters riled up, and it’s not aimed at American businesses. I’d be surprised if even a third of Republicans supported anything more than not doing business with Anthropic. Even the Nvidia shakedown got a ton of criticism and that’s just money.
AGI? My guy, it's a text predictor slot machine. Very useful tool but will never be AGI.
He said “from a company working on AGI” which is true. Not to mention that the sarcastic nature of your comment is off putting
"I can state flatly that heavier than air flying machines are impossible. — Lord Kelvin, 1895"
I'm sure this doesn't apply to you since you're not Lord Kelvin. On the other hand, people like Peter Norvig state in a popular AI textbook that, for example, they don't know why similar concepts appear close by in the vector space, so maybe you just know something other people don't.
Map problems to slot machines, guess enough slots and you're indistinguishable from GI.
I'm not taking a position here but the person you're replying to stated that Anthropic are working on AGI, not that their current LLM offering will evolve into AGI.
Ok that's different then. LLM, by definition, can't be AGI. But AGI can be AGI with another technology.
> LLM, by definition, can't be AGI.
False, and you've given no argument to the contrary. There's certainly no definition that precludes it. It isn't, currently; there's no reason it can't be, any more than there's reason that Conway's Game of Life can't be, given sufficiently interesting data to process. Any Turing-complete system could simulate AGI. It might not be the most efficient mechanism for doing so, but that's not the question at hand.
Said the biological text predictor…
Pretty rich coming from an AGI that’s running on a bowlful of mildly electrified meat. Emergent properties, my guy.
Europe doesn’t give a shit about another American company and their employees trying to dominate their markets and import their workaholic American culture. They will tell Anthropic to go home.
"Europe" is not a single entity with uniform opinions. As an European, I would much rather have hardworking people and """workaholic""" culture than regress to an underdeveloped culture fueled by laziness.
>>underdeveloped culture fueled by laziness
Which of the European cultures is "underdeveloped", exactly?
This is pretty disconnected to how EU has been behaving towards both startups and AI.
Europe doesn't care about onshoring the best AI in the world and possibly achieving AGI before everyone? That's a laughable assertion.
Not sure where you are in Europe, but in France, Macron would bend over backward.
This could kill Anthropic.
The designation says any contractor, supplier, or partner doing business with the US military can’t conduct any commercial activity with Anthropic. Well, AWS has JWCC. Microsoft has Azure Government. Google has DoD contracts. If that language is enforced broadly, then Claude gets kicked off Bedrock, Vertex, and potentially Azure… which is where all the enterprise revenue lives. Claude cannot survive on $200/mo individual powerusers. The math just doesn’t math.
None of the hyper scalers are going to stop offering Claude. All of the big 3 have invested billions of dollars into Anthropic, and have tens (if not hundreds) of billions more tied up in funding deals with them. Amazon and Google are two of the largest shareholders of Anthropic.
Anthropic is going to be fine. The DoD is going to walk this back and pretend it never happened to save face.
Tens, maybe hundreds, of billions? That’s cute. The DoD will spend $961b this year. It does that like clockwork every year, year after year.
Anthropic is not even close to too big to fail. And even if this could get settled in court 5 years from now, this can easily throw enough of a wrench into their revenue streams to kill their flywheel.
The DoD’s spend on cloud contracts is measured in single-digit-billions per year. It’s peanuts compared to the hyperscalers investments in Anthropic.
Think of it this way: each of the hyperscalers have built a handful of data centers specifically for government contracts. A handful each.
Meanwhile, AWS and GCP have dedicated over 50 new data centers solely for Anthropic to train new models, and more were announced today.
My bet is on Anthropic.
That $961 billion includes things like airplanes and bullets, tech companies are only getting a taste of that pie not anywhere close to the whole thing.
and?
The cost of a company like Amazon or Google losing their piece of that $1T annual budget is greater than their exposure to the failure of Anthropic.
Not according to published Financials.
Also $1T is dishonest. DoD spends less than 0.1% of that on cloud services.
Source?
Half of that budget gets contracted out to Lockheed, Raytheon, Northrop, Boeing, General Dynamics, etc. Those companies absolutely do spend money on the hyperscalers.
GovCloud revenue is in the tens of billions of dollars. Bedrock less so. Almost every FedRAMP product uses the same codebase for Fed and non-Fed, and this would force most FedRAMP vendors to blackball Anthropic.
The JWCC, which is larger than GovCloud, was only $9b, split across three companies, over ten years. It’s peanuts compared to the investments that the hyperscalers have with Anthropic.
JWCC is not the only project. Vendors like Crowdstrike also rely on hyperscalers to serve their products to federal customers, and the codebase is shared.
This announcement has made Anthropic toxic in the entire dependency chain because it means years of efforts and tens to hundreds of millions of dollars rearchitecting entire platforms and renegotiating contracts.
The entire cybersecurity industry has a TAM of $208 BILLION [0]
[0] - https://www.bccresearch.com/market-research/information-tech...
> because it means years of efforts and tens to hundreds of millions of dollars rearchitecting entire platforms and renegotiating contracts.
This is exactly why this announcement has not made Anthropic toxic. The entire industry knows how ridiculous this move is from Hegseth, and it’s going to be rolled back next week once the adults get back from their weekend.
I would find that a lot more plausible if people had not spent the past week giving me similar arguments, in precisely the same tone, for why this was an empty threat and would never happen in the first place. If Amazon and Google do not either bow down or immediately join a business coalition to get Trump out of power, Hegseth will be even happier to get an opportunity to prove his power by destroying them. Trump either doesn't want to stop him or has become too senile to stop him.
It is narrower than that by law, though not by their proclamation.
That label forbids contractors on DoD contracts for billing DoD for Anthropic, or including Anthropic as part of their DoD solution.
So - AWS can keep claude on bedrock, but can't provide claude to the DoD under its DoD contracts
Not entirely true.
The designation only applies to projects that touch the federal government, or software developed specifically for the federal government.
Contractors can still use Claude internally in their business, so long as it is not used in government work directly.
A complete ban would be adding Anthropic to the NDAA, which requires congress.
The DoD designation allows the DoD to make contractors certify that Anthropic is not used in the fulfillment of the government work.
The language in the tweet was
" Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic."
Is that just his fantasy or?
Example: Perhaps "Amazon US Services LLC" or whichever subsidiary they have that deals with the government will be banned from using Claude, and all of it's other subsidiaries won't?
https://www.sec.gov/Archives/edgar/data/1018724/000101872423...
Claude is in all of 'em
https://www.anthropic.com/news/claude-in-amazon-bedrock-fedr...
Well, IANAL but tweets aren't legislation. What that tweet implies is something that would have to be amended into the NDAA, which requires congress. Hegseth can't just go on a drunk rant and have everything out of his mouth become law.
The supply chain risk directive would come from existing procurement law, which only allows the DoD to require contractors to certify that Anthropic is not used in the fulfillment of any government work.
Which is also separate from Trumps' EO, which being an EO only applies to the federal government directly.
So yeah, banning any contractor, supplier, or partner from any commercial activity with Anthropic is just fantasy without going through congress first.
> Contractors can still use Claude internally in their business, so long as it is not used in government work directly.
I work in the enterprise SaaS and cybersecurity industry. There is no way to guarantee that amongst any FedRAMP vendor (which is almost every cybersecurity and enterprise SaaS or on their roadmap).
Almost all FedRAMP products I've built, launched, sold, or funded were the same build as the commerical offering, but with siloed data and network access.
This means the entire security and enterprise SaaS industry will have to shift away from Anthropic unless the DPA is invoked and management is changed.
More likely, I think the DoD/DoW and their vendors will force Anthropic to retrain a sovereign model specifically for the US Gov.
Edit: Can't reply
> This is the core assertion that is not clear nor absolute.
If Walmart can forcibly add verbiage banning AWS from it's vendors and suppliers, the US government absolutely can. At least with Walmart they will accept a segmented environment using GCP+Azure+OCI. Retraining a foundational model to be Gov compliant is a project that would cost billions.
By declaring Anthropic a supply chain risk, it will now be contractually added by everyone becuase no GRC team will allow Anthropic anywhere in a company that even remotely touches FedRAMP and it will be forcibly added into contracts.
No one can guarantee that your codebase was not touched by Claude or a product using Claude in the background, so this will be added contractually.
> This means the entire security and enterprise SaaS industry will have to shift away from Anthropic unless the DPA is invoked and management is changed.
This is the core assertion that is not clear nor absolute.
From what I’ve heard the actual restriction is just on using Claude for stuff they’re doing for the Pentagon. They’ll keep using Claude for everything else and be less effective when they work for the government, and that’s fine because everyone else working for the government will have the same handicap.
I’m sure most of their revenue is large enterprise customers who serve government with their products - this looks very bad
That's what hegseth says, but the law doesn't really say that AFAICT.
This will likely go to court, again as Dario has stated this is blatant retaliation as no US company has ever been designated a supply chain risk and they continue to operate on classified systems for 6 more months.
Yea strong odds this goes to court, the DoD’s clearly inconsistent logic is ridiculed by a judge, the designation is dropped, and everyone quietly goes about their way with the DoD continuing to use Claude according to the existing terms of the contract.
Sure, after a decade of litigation, meanwhile Anthropic goes bankrupt.
No, Anthropic could easily call their bluff.
Topics like this are where I struggle with HN philosophy. Normally avoiding politics and ideology where possible, created higher quality and more interesting discussions.
But how do you even begin to discuss that Tweet or this topic without talking about ideology and to contextualize this with other seemingly unrelated things currently going on in the US?
I genuinely don't think I'm conversationally agile enough to both discuss this topic while still able to avoid the political/ideological rabbit-hole.
You can't discuss this topic without broaching the idea that the government is acting in bad faith — that they don't actually believe that Anthropic is a supply-chain risk and that this action is meant to punish the company. But this is in the HN guidelines regarding comments:
> Please respond to the strongest plausible interpretation of what someone says, not a weaker one that's easier to criticize. Assume good faith.
If a commenter who supports the government makes the same argument that the government is making, the guidelines tell us to assume good faith.
My conclusion is that any topic where a commenter might be making a bad faith argument is outside the scope of Hacker News.
On the other hand, pretending the government is acting in good faith is probably acting in bad faith at this point.
My interpretation of that is that I’m required to assume good faith on behalf of other commenters. So, if someone makes the same argument as the government, I’m supposed to assume good faith there, but nothing requires me to assume good faith on behalf of the government. So I can say that this is obviously a shakedown without breaking the rules.
"Assume good faith" does not mean "extend an unlimited amount of good faith to demonstrably bad-faith actors".
>Assume good faith.
This is more for “assume op is not a troll” rather than “assume Donald trump never took part on Epstein’s parties”.
I’ve never taken it to apply to anything other than the interaction with other commenters.
I've been on hn for years and I see this kind of sentiment raised all the time. It is not my understanding of the guidelines.
Politics and ideology are not off topic, provided the subject matter is of interest, or "gratifying", to colleagues in the tech/start-up space.
What's important is that we don't use rhetoric, bad faith or argumentation to force our views on others. But expressing our opinions about how policy affects technology and vice versa has always been welcome, in my observation.
So, what do you think about the US government's decision, and why?
> Normally avoiding politics and ideology where possible, created higher quality and more interesting discussions.
Everything is politics and "ideology"
Being a hacker used to be an extremely political and ideological movement. Then capitalism came along and bought the term. It's about time we take that word back where it belongs.
Please at least try. There are already enough contributors here "qualified" to talk about politics.
If the last ten years have taught us anything it's that politics just isn't a topic isolated to the halls of government. It's real life. Political alignment has never so starkly indicative of your position on fundamental human morality. At the same time we've never had a government be so directly involved in private businesses.
>Topics like this are where I struggle with HN philosophy. Normally avoiding politics and ideology where possible, created higher quality and more interesting discussions.
Our whole society runs on technology. All tech is inherently political.
A "no politics" stance is merely an endorsement of the status quo.
Why would you want to be non-political in 2026? The current administration is awful in ways we couldn't have imagined. There's no sense in not talking about it.
I appreciate your restraint, and keeping this a high quality discussion space. As a political dissident myself, I don't mind some threads going political, I expect them to. The best ones are when there is a lot of disagreement or debate. As long as its not in every unrelated thread....
In theory, this is why there should be competition in industry, because it removes the capability of a single large actor to be able to control the government's access to things.
Oddly, though, it seems like that should solve this problem as well. I'm not sure why the Department of Defense insists on Anthropic's models in particular; one would think one of the other players, at the very least least xAI, would be willing to step in and provide the capability Anthropic doesn't want to provide.
Everyone is getting wrapped around the axel here but this is about the big picture, not the specifics. A private company should not have the ability to dictate how its technology is used by the government. If they can’t agree to that, then don’t sell your technology to the government. Personally, I don’t want to be spied on by the government with it (I don’t think their tech does that) but I also don’t want Anthropic having operational control over a mission.
McCarthyism began in 1947, with Truman demanding goverment employees be "screened for loyalty". They wanted to remove anyone who was a member of an "organization" they didn't like. It began with hearings, and then blacklists, and then arrests and prison sentences. It lasted until 1959. (https://en.wikipedia.org/wiki/McCarthyism)
This is the new McCarthyism. Do what the administration says, or you will be blacklisted, or worse.
Feels a bit like Jack Ma and Alibaba
But McCarthy was right
> Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic.
I’m sure the lawyers just got paged, but does this mean the hyperscalers (AWS, GCP) can’t resell Claude anymore to US companies that aren’t doing business with the DoD? That’s rough.
Probably yes. Additionally the (probably more for AWS) won't be allowed to use it internally either. This will probably apply to all the top SaaS/software companies unilaterally.
Additionally, every major university will undoubtedly have to terminate the use of Claude. First on the list will be universities that run labs under DOD contracts (e.g. MIT, Princeton, JHU), DOE contracts (Stanford, University of California, UChicago, Texas A&M, etc...), NSF facilities (UIUC, Arizona, CMU/Pitt, Purdue), NASA (Caltech).
Following that it will be just those who accept DOD/DOE/NSF grants.
There is no evidence that what you say is true. A tweet is not a legally binding statement.
When did legality apply to this administration?
What part? Are you doubting that they are being designated as a supply chain risk? Or the implications of being designated as one?
We do have a recent example with Huawei, and it did fall just like this - and that was just some hardware.
It will be true as soon as it becomes official though, assuming they actually go through with it and this is not just a bargaining tactic.
Won’t that require an act of congress? How likely does that seem?
Huawei was not on the NDAA (the congress part) until August 2019, well after companies started cutting ties in April/May of that year
Billable hours will win figuring it out but in theory, no because they can’t test it or use it.
Generally any machine that touches Supply chain Risk software cannot ship any software to DoD. AWS has separate clouds but software comes from same place.
Bigger question is whether government contractors can use any Open Source software after this. Open Source is a big part of the supply chain.
It means everyone waits for the injunctions.
(edit: I'm most likely wrong)
You got it backwards, can't use claude if you ARE doing business with DoD.
Presumably AWS/GCP don't care, its up to the end customer to comply. Not like GCP KYC asks if you work with DoD.
AWS/GCP/Azure all do business with the DoD and at least AWS and Azure use Claude a decent amount internally. AWS’s Kiro tool (which is used internally instead of Claude Code) relies entirely on Claude models.
This is almost certainly going to be rolled back, because I guarantee the DoD isn’t going to stop doing business with the hyper scalers, and the hyper scalers aren’t going to stop doing business with Anthropic.
I don't think he got it backwards, at least if Hegseth's statement is accurate. AWS, GCP, etc. all do business with DoD. If they, as DoD contractors, are no longer allowed to do business with Anthropic, then presumably they have to stop re-selling or hosting Anthropic's models to anyone.
Ah, true. Well then, what makes GCP/AWS more money? DoD contracts or Claude resell fees? They could drop DoD though I guess I see how this will go...
>> at least if Hegseth's statement is accurate
Oh you tender babes, trying to logic the meaning of what the lieutenant of the biggest crime syndicate in the world means with his words, as if this was a well thought-out strategy... it's a shakedown; it would make more sense to ask "at least if Hegseth is sober..."
If I had to bet, there will be some kind of face-saving climbdown by the end of next week. But all I can do right now is read the words on the page.
So GitHub Copilot will remove Anthropic as an LLM provider, I suppose?
Agree with other reply. I don’t think it’s backward. No they said any commercial activity. Does not feel like a stretch that commercial activity includes reselling api usage.
have you tried punching in "Huawei" the shopping portal on google.com in the US?
No, what happens when one does?
nothing, which is the point though
Even more extreme, that might mean they won't be able to offer Claude to non-US companies at all.
I don't see how you get that reading. Anthropic is clearly allowed to sell Claude to companies not doing business with the US Military. If anything that's more likely to be non-US companies.
IIRC, the supply chain risk designation is sticky which is why it tends to ultimately mean "nobody can work with this". Amazon using claude means a DoD company can't use Amazon. Every business that touches claude gets tainted.
It's a bit like how the US Cuba sanctions worked and why they effectively isolated Cuba from everything.
Because Anthropic sells Claude through other companies that in turn do business both with Anthropic and the government. These intermediaries, large cloud companies, can't offer Claude anymore if they want to keep the government as a customer.
The government is faaaaaaaaaaaar too invested in Azure and AWS for Microsoft or Amazon to give even half a shit. The DOD has no where else to go and the companies know it. They'll sit on their hands until the legal maneuvers play out, which will take longer than this administration will be in office.
You expect hyperscalers to play chicken with the DoD?
The courts have historically been pretty consistent about giving the DoD whatever the fuck they want, going back to WW2 and even longer for the predecessors of the DoD. I agree that the next administration might reverse it, but the thing is, the government will stay irrational longer than Anthropic will remain solvent.
The US government told every American company to stop doing business with Huawei and they all did it overnight, even when it cost them billions. TSMC stopped fabricating for them, Google pulled Android licensing… The machinery of sanctions compliance is extremely well-oiled and companies fold instantly because the outcome of noncompliance is literally getting thrown in prison.
This is also true, unless the government can force them to drop Anthropic on the basis that the alternative- the government dropping them- is unworkable.
Or Pete Hegseth will threaten to do the same to them unless they comply, and they will demonstrate the same inexcusable cowardice the American business class has consistently demonstrated this past year. Hope I'm wrong and this has finally woken them up!
Sorry, the "they" referred to the hyperscalers
There is no way they can just stop selling Opus 4.6. This will crater the market.
This doesn’t erase Claude, and even if it did Gemini and Codex are there to replace it.
Even if a ton of companies have to switch over to an alternative, it won’t be catastrophic to the economy.
The stock market will be spooked if the US govt can willy nilly high trajectory darling of the AI world like this though.
Who's next? OpenAI? Google? What if they refuse to allow the DoD to use AI with zero safeguards and Trump's goons decide they are also a "supply chain risk"?
No. The stock market has understood for generations that it's the guys with the guns that protect their gold. The stock market will have a sigh of relief.
Wait, what about Bun?
"They have threatened to remove us from their systems if we maintain these safeguards; they have also threatened to designate us a “supply chain risk”—a label reserved for US adversaries, never before applied to an American company—and to invoke the Defense Production Act to force the safeguards’ removal. These latter two threats are inherently contradictory: one labels us a security risk; the other labels Claude as essential to national security." from Dario's statement (https://www.anthropic.com/news/statement-department-of-war)
Supply chain risk ? Seems the risk here is the US Gov't wanting free reign to do whatever they want - - when they want.
Look no further than the famous expose by Mark Klein, the former AT&T technician and whistleblower who exposed the NSA's mass surveillance program in 2006, revealing the existence of "Room 641A" in San Francisco. He discovered that AT&T was using a "splitter" to copy and divert internet traffic to the NSA, proving the government was monitoring massive amounts of domestic communication.
And I think on big difference between <2006 and now is that back then nobody knew about it - now they just request it in public.
I served on the eboard of CWA local 9410 when all of that was going down.
Words cannot describe how crazy things were at that time.
I feel like someone will make a movie about it someday.
The risk is a business that doesn't lick the boot might speak truth to power.
The real question we should be asking is what others HAVE agreed to. Has OpenAI just agreed to let the government go crazy with their models?
If you read Anthropic statement carefully, they explicitly confirm they are already working with the U.S. government on a range of military and national security use cases, many including areas that clearly relate to real world lethal operations.
They are only refusing two narrow, but important categories. Framing this as blanket "refusal to support the DoD" feels like an angry, reactive own goal rather than a careful reading of what they actually said.
So far the march toward dictatorship keep being detoured by sheer incompetence. In any case, is hard to seize power when you can’t organize a group chat...
Basically now all those projects are screwed and need to restart with another provider. I'm sure that's not going to be a massive PITA and delay for all involved.
Elon has agreed to all demands and can’t wait for gigahitler to take the reigns. I swear there is no room for good guys in this is there.
The military already has access to Grok, but doesn't want it, because it's an inferior model, even compared to open source ones. So the military would probably choose to replace supply chain risk Claude with Qwen or Kimi before Grok.
It would be untouchable irony for the US to cut all ties with Anthropic and replace them with models developed by Chinese labs. The Onion becomes more irrelevant with each passing day.
How many generations does it take before the historians/archeologists uncover old issues of The Onion and decide it was the authoritative news of the day?
I thought I had a sense of dejavu. I was wrong.
Grok is according to most benchmarks pretty close to SOTA. It is where the leaders were just a few weeks ago.
Which exactly is best changes on almost a weekly basis as different companies tweak their best model. I doubt the military would want to be switching supplier every week.
I think that tells you more about the uselessness of SOTA benchmarks.
Musk was embedded in the military industrial complex with Thiel since day 1.
https://www.mintpressnews.com/pentagon-recruiting-elon-musk-...
Rumor has it they like to tickle each others' homunculi right in the region known anatomically as the inferiority-superiority complex.
> Altman says OpenAI agrees with Anthropic’s red lines in Pentagon dispute
https://thehill.com/policy/technology/5758898-altman-backs-a...
He's probably lying. Or he "agrees" but will cross the line anyway.
Altman is an Aes Sedai. He speaks no word that is untrue, but is one often most deceptive people I’ve ever heard.
This is only because Altman knew he’d already lost this business to Musk.
Can someone in plain terms explain what this is really about?
Anyone can use Claude afaik?
From the public comments over the last few days, my guess is they want a militarized version of Claude. Starting with a box they want to put in the basement of the Pentagon where Antropic can't just switch off the ai. Then some guardrails are probably quite bothersome for the military and they want them removed. Concretely if you try to vibe-target your ICBMs Claude is hopefully telling you that that's a bad idea.
Now, my guess is in the ensuing lawsuit Antropic's defense will be that that is just not a product they offer, somewhat akin to ordering Ford to build a tank variant of the F150.
> Concretely if you try to vibe-target your ICBMs Claude is hopefully telling you that that's a bad idea.
On the non-nuclear battlefield, I expect that the goverment wants Claude to green-light attacks on targets that may actually be non-combatants. Such targets might be military but with a risk of being civilian, or they could be civilians that the government wants to target but can't legally attack.
Humans in the loop would get court-martialed or accused of war crimes for making such targeting calls. But by delegating to AI, the government gets to achieve their policy goals while avoiding having any humans be held accountable for them.
I used to not be big on conspiracy theories. But I'm going to give this a shot because many of the old ones turned out to be true.
Why can't Grok achieve this? Everyone is saying they don't want to work with Grok because Grok sucks, but it's good enough for generating plausible deniability, isn't it?
[delayed]
> Starting with a box they want to put in the basement of the Pentagon where Antropic can't just switch off the ai.
They already have that. By definition. If Anthropic has done the work to be able to run on classified networks, then it's already running air-gapped and is not under Anthropic's control.
The thing is, just because you're in a SCIF doesn't (1) mean you can just break laws and (2) Anthropic don't have to support "off-label" applications.
So this is not about what they have and what it can do today - it's about strong-arming anthropic into supporting a bunch of new applications Anthropic don't want to support (and in turn, which Anthropic or it's engineers could then be held legally liable for when a problem happens).
>akin to ordering Ford to build a tank variant of the F150.
It worked for Porsche ¯\_(ツ)_/¯
Best summary by far that I've seen:
https://www.astralcodexten.com/p/the-pentagon-threatens-anth...
Discussed here:
https://news.ycombinator.com/item?id=47154983
Claude won't answer questions about what cities you should nuke in what order. The Pentagon wants Claude to answer those sorts of questions for them.
Edit: oops, I misunderstood. This seems to be more about contractual restrictions.
Claude will answer all of those questions. The restriction Anthropic has is letting Claude pull the trigger and vibe-murder with no humans in the loop.
This restriction is apparently "radically woke"
I reached to answer but idk what you mean by the second question. Long story short, Department of “War” wants Anthropic to say theres no restrictions on their use of Claude, Anthropic wants to say you can’t use Claude for domestic mass surveillance or automating killing people domestically or in foreign countries. Rest is just complication. And don’t peer too closely at the “Do”W”” wants Anthropic to say $X, the Team Red line (or, whatever’s left of them publicly after this last year) is basically “you can’t tell the gov’t what it can and can’t do, that’s it, it’s not that Do”W” will use it for that”
They want Claude to process tasks like "identify the terrorists in this photo" and "steer this drone towards the terrorists" — Anthropic refused.
top signal
> Can someone in plain terms explain what this is really about?
This administration built almost entirely of dunces and conmen has convinced itself/been convinced that chatbots will help them in deciding where to send nukes, and/or they are invested in the incredibly over-leveraged companies engaged in the AI-boom and stand to profit directly by siphoning taxpayer dollars to said companies. My money is on the latter more than the former, but they're also incredibly stupid, so who's to say, maybe they actually think Claude can give strategic points.
The Republicans have abandoned any pretense of actual governance in favor of pulling the copper out of the White House walls to sell as they will have an extremely hard time winning any election ever again since after decades of crowing about the cabal of pedophiles that run the world, we now know not only how true that actually is, but that the vast majority are Conservatives and their billionaire buddies, and the entire foundation and financial backing of what's now called the alt-Right, with some liberals in there for flavor too of course.
If this shit was going down in France, the entire capital would have been burned to the ground twice over by now.
> they will have an extremely hard time winning any election ever again
Heard that one before. We'll get a reprieve of 4-8 years and the vote will go to the fascists again. Take that to the bank.
A girl can dream.
Or there won't be another election. They keep telling us there won't be another election. Why aren't we more alarmed by that? Why are we assuming they are lying about that?
I prefer to call them chatboxes. It's appropriately belittling. The department of killing wants their chatbox to tell them who to kill.
> If this shit was going down in France
your view of France is severely outdated
Yes. All companies that deal with the government have agreed to let the government do whatever it wants within the bounds of whatever it is those companies do.
Probably just gonna go all in on MechaHitler!
It's scary to me that there are people out there who don't see this kind of zero-integrity behavior as disqualifying in elected officials.
That's a shame. They might at least continue to work together to spy on foreigners. I don't understand the fuss anyway, what do claude models do that gpt and gemini can't?
As a foreigner, i see this as a great thing! I was about to cancel my Claude sub, but now i might hold on to it for a little and see how this plays out.
For these people, it is just about control.
Future Trump rally: "And I hear Anthropic monkeyed with their dishonest chatbot Claude. They turned it Democrat! They trained it to say we lost the election against Sleepy Joe!"
it's more the way they do them.. you've used them right?
Sure but I don't find them irreplaceable. Actually anthropic models have dropped out of my top ten usage this month. I only use opus occasionally for writing plans, its been pretty unreliable at executing.
Is this the same Administration that reversed a previous block, and allowed NVIDIA to sell H200 to China?
Well, you see, that's completely different. Nvidia agreed to give them money!
Silly me...its true!
- $1,000,000 donation from NVIDIA CORPORATION to the Trump–Vance Inaugural Committee.
- $1,000,000-per-head Mar-a-Lago dinner where Nvidia CEO Jensen Huang attended.
- Jensen Huang’s contribution toward Trump’s "White House ballroom" project. Confirmed, but undisclosed value...lets says at least another $1,000,000?
>> Well, you see, that's completely different. Nvidia agreed to give them money!
Also I believe NVIDIA's supposed to pay the US government 15% of its revenues from Chinese sales:
https://www.ft.com/content/cd1a0729-a8ab-41e1-a4d2-8907f4c01...
Which is incredibility short term thinking. You're in strategic competition, and you compromise you position for a bit of cash?
No one has ever accused Trump of being in this for the long term strategic vision lol
$1,000,000 doesn't seem like a lot of money for them, why would it matter to them?
A good reason to outlaw bribes is that politicians tend to be incredibly cheap and offer an extremely high ROI. Albeit at the cost of a nice democracy.
Ghengis Khan didn’t need your chest of gold, he owned many gold mines. Regardless, he was going to take it from you the easy way or the hard way.
You're forgetting that this is the same guy who managed to bankrupt a casino. He's not actually that good with money and until the latest bribe channels opened, eg Trump Coin and the Board of Peace, opened their finances may have been in a bit of a mess. Also I'd bet the ballroom donation was much larger, it's a massive blackhole of graft waiting to happen.
It's also not solely about money, you can get far just knowing how to chum it up with Trump when you get in the room with him. Look at the odd quasi-bromance between him and Mamdani who you'd expect to be enemy #1 but Mamdani knows how to schmooze the exact type of New York Guy Trump is.
Ahem, depending on how you count, he bankrupted 4-6 casinos.
To Nvidia, or to the recipients?
Both?
For fascism, it's not always about getting something you think is a lot. It's about a power relationship. Trump has demonstrated that Nvidia will bow to his will.
It's also potentially an implementation of the foot-in-the-door technique (https://www.simplypsychology.org/compliance.html). It's a common manipulative strategy where you get someone to do a small favor for you which makes them much more likely to do a large favor for you later.
Ah yes, again the: "I am so rich I could not possibly be corrupt!"
"Trump’s Profiteering Hits $4 Billion" - https://www.newyorker.com/news/a-reporter-at-large/trumps-pr...
"How much money President Trump and his family have made" - https://www.npr.org/2026/01/14/nx-s1-5677024/trump-profits-m...
Good thing this administration will be a lame duck in 8 months, and they know it.
"trump is definitely gonna lose the election" is a prediction I've heard many times. I know better than to trust it by now
At least twice. Luckily, that's the max number
That's part of why they are trying to take control of elections, which have (I believe) historically been the responsibility of each state.
a very optimistic view
https://kalshi.com/markets/controlh/house-winner/controlh-20...
https://polymarket.com/event/which-party-will-win-the-house-...
The branch of government tasked to execute the law has been ignoring laws. So we'll get a (from Trump's point of view) adversarial congress, so what, let's ignore them, what are they going to do about it?
Looking forward to a military platoon defying orders and seizing the president, hey, all countries suffer through coups, about time this young democracy go through one!
> about time this young democracy go through one!
Did you skip class they day that discussed the Civil War?
The terms of these markets do not account for a scenario, quite likely if authoritarian takeover does happen, where the House of Representatives is a rump organization which does not exercise effective power. There was a years-long period in Venezuela where the country's traditional legislature met and conducted business under the leadership of the opposition party, but actual legislative power was held jointly by the Supreme Court and a secondary legislature that Nicolas Maduro set up.
So cool we can bet on whether the Trump admin will attempt another coup - what a time to be alive
Are you sure? They have one skill: playing social media, and it serves them well.
Unless ICE ensures it’s is a ”fair” election with the ”correct” outcome.
Luckily, the oval office is on the ground floor, so it's safe to stand next to the windows
Zombie Duck
The Purpose of a System is WHAT IT DOES!
https://www.astralcodexten.com/p/come-on-obviously-the-purpo...
I feel this is a facile interpretation of the phrase.
A more steel-man interpretation of POSIWID--the way it's intended to be understood--would be: "What an established system accomplishes in the long term reveals something important of the the true preferences of the various interests that control it, which can easily diverge from the system's stated goals."
Great read. I've always noticed that the type of argument invoked is often less telling than when and in which context you invoke that argument.
You can make a lot of claims and they can match to reality a lot - normally people think of evaluating things in terms of a strict "does this fit or does this not", but it's often the meta-style (why do you keep bringing up that argument in that context?) that's important, even if it's not "logically bulletproof".
https://www.astralcodexten.com/p/highlights-from-the-comment...
Wow that post is bad. The author clearly never actually attempted to understand what POSWID actually means and where it is coming from. Perhaps, instead of looking at Twitter, they should have opened Wikipedia. Or, better yet, Stafford Beers books (though admittedly, he was a pretty atrocious writer).
The follow-up is slightly better. But still not very convincing, IMO. They get far too stuck on a literal interpretation. Of something that self-describes as a heuristic.
> what POSWID actually means
The phrase does not make more sense even if we go all the way back to Beers. I certainly don't feel alone in not understanding how he went from his (fair) observation that "[There's] no point in claiming that the purpose of a system is to do what it constantly fails to do" to his more controversial conclusion: "The purpose of a system is what it does (aka POSIWID)".
Surely, there were many more sensible (but perhaps less quippy) stops between the two.
Unconstitutionally, no less:
"No Tax or Duty shall be laid on Articles exported from any State.".
I would not be surprised if an outcome of this may be a 10% government stake (maybe golden share owned by Trump) in Anthropic.
It feels like when you are negotiating a contract for job with a toxic employer who you still don’t know they are toxic yet.
Trump wrote a long rant on Truth Social and ordered ALL federal agencies to stop using Anthropic. Not just the department of defense. This is straight up authoritarian.
Meanwhile, irrelevant "AI Czar" David Sacks, member of the PayPal mafia alongside known Epstein affiliates Elon Musk and Peter Thiel, is furiously retweeting all the posts from Trump, Hegseth, and other accounts. He is such a coward and anti American:
https://xcancel.com/davidsacks
I don’t see a contradiction here. If control is out of the hands of decision makers, that’s a supply chain risk . Were it not for that, the service is seen as critical to national security.
I dunno, safeguard seems like a weasel word here. It’s just reserving control to one party over another. It’s understandable why the DoD(W) wouldn’t like that.
I'm pretty sure you (and others) are trying to apply some kind of guess at the "supply chain risk" designation, but it means something specific.
Here's the term defined in an official context:
https://www.acquisition.gov/dfars/252.239-7018-supply-chain-....
Since the link is still broken, I tried encoding the final dot as %2E
https://www.acquisition.gov/dfars/252.239-7018-supply-chain-...
That link is broken for me but I assume you meant to link to [0]. I think if there is a “safeguard” in a system, that definitely fits the bill of a supply chain risk. The only vague term here is “adversary”.
[0]: https://www.acquisition.gov/dfars/252.239-7018-supply-chain-....
Working link: https://www.acquisition.gov/dfars/252.239-7018-supply-chain-...
HN separates trailing dots from URLs, so that you can have working URLs at the end of a sentence. Hence you have to percent-encode trailing dots if they are a necessary part of the actual URL. (Same for some other punctuation characters, probably.)
This behavior is common for auto-hyperlinking of URLs in running text, so it’s bad practice to have such URLs.
Ugh, sorry for the broken link, I even pasted the same string into a new tab to make sure it worked because I thought the period at the end looked weird, and it was fine. Dunno how it got mangled.
[EDIT] Oh man, yours is like that too? WTF.
[EDIT2] If I follow your link, hit the 404 page, then add a period at the end of the URL, it does load. God that's strange.
Well, we ended up on the same page in any case, in at least one sense.
Yes, we both accurately located and linked to the "page not found" page.
That gave me a good, actual LOL, thanks for that one.
Did you edit it to fix it? Is HN refusing to include the period as part of the URL?
404
"Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic."
This is authoritarian behavior. You're having trouble negotiating a contract, so instead of just canceling it - you basically ban all of F500 from doing business with that firm.
I think it’s sovereign behavior and what’s the point of being sovereign if you don’t exercise the power of the sovereign?
I guess I would support the democratically elected sovereign over the private corporation.
What does being sovereign have to do with anything in this case?
The sovereign is the ultimate authority. In the USA “we the people” delegate the sovereign power to our elected officials.
They are now exercising that power in the interest of the people (they believe) that grant that power.
You don't actually believe in the core tenets of the USA if you think that the government should have or should exercise unchecked, abusive power.
American people: latinamerican here. Maybe it's silly to root for a country in the world hegemony arena. I've usually been partial to the USA over China. Now I'm not rooting for your country anymore. As far as I'm concerned, I'd rather have China being the foremost power, at least they seem to be less keen on invading or heavily strong-arming latinamerica
and USA created Islamic terrorism that is plaguing the whole world
Surely you mean Zionist terrorism.
I empathize, but surely China is not the right choice? Can we please have like, Australia? Or a unified EU?
Hard decision by Anthropic, but at least they can sleep well at night knowing their products doesn’t kill human beings around the world.
That’s the crazy thing. This whole dispute was over Anthropic saying no to fully automated kill bots. They only required there be a human in the loop to press the button.
Anthropic didn't even say "no", it was more of a "not yet, let's work on this".
I really wonder what Palantir's role in all this is because domestic surveillance sounds exactly like Palantir and whatever happened during the Maduro raid led to Anthropic asking Palantir questions which the news reports is the snowball that escalated to this.
They also said no to fully automated AI domestic surveillance. I suppose non-US citizens like me are screwed but that's at least some small comfort for the natives. FVEY will just spy on each other and share but at least someone tried.
There were two red lines, as I understand it -- first, automated kill bots, and second, mass surveillance.
Mass domestic surveillance of American citizens (they were OK with surveillance of other countries).
No. There was only one red line.
Bend over and take or not.
Neither of those red lines should be controversial. What American citizen thinks terminators and Big Brother are desirable?
MAGA (as long as the terminators are pointed towards the other side)
Citizen 1?
The ones that still assume big brother will be spying on and killing the people they hate. Trump openly campaigned on getting revenge on his enemies. I can only assume his supporters want this. The danger of course is if/when the leopards eat their faces
I guess the problem for Trump is if he orders the army to gun down protesters, there’s a good chance they will refuse to do it. While a bot can just be prompted to go ahead.
This one here is the future I am most scared of.
Yeah, but imagine if it were true
I think it’s far more likely this is about the other sticking point- using it to spy on US citizens.
If we were able to give the Ukrainians fully automated kill bots, and those kill bots enabled Ukraine to swiftly expel the Russians from their territories, would that not be a good thing? Or would you rather the meat grinder continue to destroy Ukraine's young men to satisfy some moral purity threshold?
If we could give Taiwan killbots that would ensure China could never invade, or at least could never occupy Taiwan, would that be good or bad? I have a feeling I know what the Taiwanese would say.
While we're at it, should we also strip out all the machine learning/AI driven targeting systems from weapons? We might feel good about it, but I would bet my life savings that our future adversaries will not do the same.
You seem to see everything from a binary perspective. China bad, Taiwan good. Russia bad, Ukraine good.
The world is more nuanced than that.
But to answer your question. No we should not give anyone automatic kill bots. Automatic kill bots shouldn’t even be a thing.
Yes, I think Russia's invasion of Ukraine is quite clearly a binary Russia=bad, Ukraine=good. Same for the impending Chinese invasion of Taiwan. Perhaps you could explain the nuances under which Russia was the good guy? Better yet, maybe you could explain it to the Ukrainians who have been displaced, or the family members of those who have been killed, or the soldiers who have been permanently maimed?
Whether you or I like it or not, automatic kill bots will be a thing. It will only be a question of which countries have them and which do not.
And there is evidence automated killbots were already used in Gaza (not that that's a good thing).
Generally, in war, there are no rules, and someone is going to make automated killbots, and I expect one place to see them quite soon is in the Russia-Ukraine war. And yes, I'm hoping the good guys use them and win over the bad guys. And yes, there are good guys and bad guys in that conflict.
Ukrainian young (24 y.o.) man here. Living and working in police 30 kilometres away from the actual frontline.
No, thanks, we don't need those "fully automated kill bots". There's absolutely no guarantee that they wouldn't kill the operator (I mean, the one who directs them) or human ally.
We're pretty much fine with drone technology we have.
But for me personally, that's not the most important point. What is more important - and what almost no one in the Western countries seems to realise (no offence, but many of westerners seem to be kind of binary-minded: it's either 0xFFFFFF or 0x000000, no middle ground at all) - is that on the Russian side, soldiers are not "fully automated kill bots" either. Sure, there's a lot of... let's say - war criminals. Yes, for sure. But en masse they are the same young men that you can see on the Ukrainian side. Moreover, many people in Ukraine have relatives in Russia, and there already were the cases where two siblings were in different armies, literally fighting with each other. So in my opinion, "fully automated kill bots" are not an option here. At least unless you deploy them in Moscow and St. Peterburg to neutralize all of the Russian elites, military commandment and other decision-making persons of the current regime.
The thing about building fulling automated kill bots is then you've built fully automated kill bots.
Fully automated kill bots are coming, whether any of us like it or not. The question is, which militaries will have them, and which militaries will be sitting ducks? China is pursuing autonomous weapons at full speed.
Personally, I think it'd be great to have the Anthropic people at the table in the creation of such horrors, if only to help curb the excesses and incompetencies of other potential offerings.
'yet'. Their reason for not allowing autonomous weapons usage was it isn't ready, not that they wouldn't do it on principle. Only the surveillance objection was on principle.
A bit of a cop-out, don't you think?
They still pay taxes, which fund the US government, which kills innocent human beings around the world...
I don't think it was that hard because if they had caved a LOT of employees would have quit.
Sleep well in a box under the overpass maybe. If Amazon can’t serve Anthropics model until the courts get everything figured out it will be too late for them.
I’m just laughing at the possibility of it he US military being forced to use Chinese open source AI models because every US model provider refuses to work with them.
>because every US model provider refuses to work with them
Zero percent chance of that happening as long as xAI exists.
Would be even funnier if they still chose Qwen over Grok.
WW3: Chinese army of intelligent bipeds vs USA waifu memes and based jokes.
Could the NSA use a national security letter to get a copy of a major private LLM?
a letter??
https://en.wikipedia.org/wiki/National_security_letter
Pete Hegseth is frantically asking Deepseek to come up with targets in Iran and some plausible objectives he can sell to the public.
As written this would be the end of Anthropic. AWS, Microsoft et al are all suppliers of the DoW and as written they must immediate stop doing business with Anthropic. Will be interesting to see how this unfolds.
TACO
Strange times. I truly feel these are the last days of our Republic. Especially if more aren't willing to take a stand.
As a Canadian looking in, I see people talking about a 36% approval as low.
How is it that high!?
That means that more than 1-in-3 of your countrymen are ride-or-die, and it's just heartbreaking to see that we're going to have to launch that many people into the sun.
To counter point, do you think AI companies located on our adversaries turf will take the same stand? I agree its nightmarish to think of AI surveillance. But why is that being lumped in with weaponry? I see these as two separate issues.
Do we need a "human in the loop" when targeting autonomous machines?
Anthropic isn't even taking a particular hard stance. Their mass surveillance prohibition only applies to domestic spying, so they're a-OK with spying adversaries. If all of the AI companies all over the world took the same stance, it wouldn't improve the life of Americans one bit.
The only other thing that the foreign AI companies could do is say no to automated killing bots, which doesn't even seem like that good of an idea considering that your countrymen will most likely have to interact with these robots that can kill without any oversight.
> "I see these as two separate issues."
... in the same sense as the two sides of a coin are separate sides maybe.
I'd say you're right, except that Trump is near death (maybe) and (more importantly) very unpopular.
He's more polarizing than usual maybe with stronger approve/unapprove ratings but his net popularity is in line with most 2nd term presidents at this stage.
https://www.realclearpolling.com/polls/approval/trump-obama-...
Yes, not historically unpopular, but enough that "I'm running for a third term" will be hard to pull-off.
Remember to vote in this year's midterms (Nov 3) if you're eligible. I don't think it's off-topic.
If anything, isn’t this admitting that the government thinks Anthropic has better technology than OpenAI, Grok, etc?
Sam Altman says OpenAI shares Anthropic's red lines in Pentagon fight [1]
So OpenAI will also be marked as a supply chain risk too, right?
[1]: https://www.axios.com/2026/02/27/altman-openai-anthropic-pen...
Really hoping for an official statement from oai. If all large llms are a supply risk, I guess it's a crash
Should military contractors put conditions on the use of their weapons? Here's our tank, but you can't invade Iran with it? We think your invasion of Venezuela is illegal, we're activating the kill switch on your jets. That's a real dangerous proposition.
They can, but the government can always just not buy their stuff.
That's not what the government is doing here.
If the T&C is agreed to up front, why shouldn't they be able to? If their client or potential client doesn't like the T&C, they can find another vendor.
This is good news all around, especially with OpenAI's statement siding with Anthropic.
Anthropic folks: I've been a bit salty on HN about bugs in Claude Code, but I feeling pretty warm and fuzzy about sending you my cash this month.
https://xcancel.com/i/status/2027507717469049070
Why does everyone associated with this administration sound like a 17 year old who got dumped when they post on twitter.
Basically a reflection of the average intelligence in the U.S.
Because this administration is entirely composed of those same 17 year olds, older but not any more mature.
Ukrainians and Russians are experimenting with FPV drones using AI for target acquisition and homing. Not yet economically viable because it is cheaper to give your FPV fiber spool instead of Nvidia Jetson to bypass jamming.
When we have first politician blown to bits by autonomous AI FPV there will be sheer panic of every politician in the world to put the genie back into the bottle. It will be too late at that point.
Anthropic is correct with its no killbot rule.
Autonomous loitering munitions with 'AI' (image classification CNNs) are already in service and have been used - most demonstrably by the IDF.
Even during the Nagorno-Karabakh war, Azeri loitering munitions were able to suppress Armenian air defenses by hitting them when they rolled out of of concealment. I believe that killchain requires a level of autonomous functionality.
Azerbaijan was buying a lot of weapons from Israel prior to Nagorno Karabach war, so it is very likely that you have been talking about same weapon system in both cases.
However Russians and Ukrainians are using AI recognition in recon drones, but not yet in FPV. There is strong suspicion that long range one way attack drones are using AI during terminal guidance, but I did not see it confirmed by either side.
Decades of speculative science fiction, thought experiments, and discourse led to this. It’s gratifying to see that we’ve garnered enough concern, a major AI lab risking this to reign in the potential of runaway AI disasters. Hopefully we see other labs follow.
Recent and related:
Statement from Dario Amodei on our discussions with the Department of War - https://news.ycombinator.com/item?id=47173121 - Feb 2026 (1508 comments)
There is clearly a need to codify into all of these historical acts that they can't be invoked unless there is a declaration of war (or some other appropriate prerequisite).
This administration consistently exploits what were designed to be emergency powers because no such requirement exists. Leave no room for interpretation.
The current administration scoffs at laws. Nothing stopping them in that case from declaring war on Nauru and doing all the same. The solution is a sane, informed electorate, which is much more difficult in this age where a few disgustingly rich people have so much influence over news and media.
It seems like some comments here are from merged threads AND front-dated?
Makes for very confusing reading when comments from "1 hour ago" are actually on preceding events from earlier, before TFA news (announcement of designation).
mods: Especially in sensitive and rapidly developing situations like this, please don't mess with timestamps of comments. It's effectively revisionism.
This sounds like a message to would-be founders: don't base your company in the US. The strongest markets to do business are the ones with the most freedom from government meddling. In the US, big government is happy to use its power to crush private enterprise that it doesn't like.
Note that previously this label has been applied (nearly?) exclusively to non-US companies. US companies that don't do business with the DoD are not affected, and non-US companies that do business with the DoD are affected.
Name one truly major market that is more business friendly
Singapore? The UK, apparently, since they don't do these things?
I think the argument would be that the US is rapidly becoming un-business friendly in the same way that Russia is.
It's nice to see Anthropic sticking to their terms. I just have one question in all this. Why is Anthropic being singled out when it seems all the other big players are down to play with the DoD? Is this just a pissing match, or have the Anthropic models been proven the real winner for them?
It's same reason this administration recently tried to indict six Congresspersons for urging military members to resist "illegal orders." They want to demonize anyone who isn't blindly loyal to their side.
It may not be obvious. But this is actually a good thing when we looking back in a few years. I always feel weird that executive branch can just destroy private enterprise with "Supply-chain Risk" / "Terrorist List" without Due Process.
I guess the worry is that we don't get Due Process here and they destroy them to make an example of them.
That's a good thing right? In a capitalist society, you cannot just burn $300B without consequences. Not to mention it is not just anyone's money. It is Saudi's.
It's basically legal hacking.
Hacking is using a system in a way it was not intended to be used.
Here it is that, but applied to the law.
Hegseth and friends are a bunch of black hat legal hackers.
Labeling a company that refused to comply with nakedly authoritarian orders is a true New Speak moment
> "Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic."
Does this mean Azure & AWS will have to stop offering Claude as a model?
I wonder, can't Amazon create a new legal entity to split AWS into "AWS-for-DoD" and "AWS-for-everyone-else"? So one can work with Anthropic and the other can't. Not sure how it works in the US.
Good PR for Anthropic: the DoD already has contracts with OpenAI and xAI, but is still so eager to use Claude that they must threaten Anthropic.
How many layers deep does this go? Does Microsoft using Claude to develop their Word products mean the US government has to switch to linux?
"Department of War" - I suppose one could give them credit for being honest but what bastards...
The name is the Department of Defense. Congress did not vote to rename it, so the name hasn’t changed.
A drunkard, ex-fox news host, wants mass surveillance and automated killing, what could go wrong?
I wish I thought enough Americans had the spine required to stand up to this, and I know for a fact that a lot do... the solution is literally written into your constitution.
Does Anthropic have standing to sue to Government for libel? I don’t think the Government is allowed to arbitrarily designate a company a supply chain risk without good cause.
I already loved Claude models, and this makes me even more eager to use them.
Help me understand the line Anthropic is drawing in the sand?
Don't get me wrong i'm glad they are unwilling to do certain things...
but to me it also seems a little ironic that Anthropic literally is partnered with Palantir which already mass surveills the US. Claude was used in the operation in Venezuala.
Their line not to cross seems absurdly thin?
Or there is something mega scary thats already much worse they were asked to do which we dont know about I guess.
The whole reason this is happening is because Anthropic looked into how Claude was used in the Maduro op and found it to violate the negotiated terms of service.
Their hard lines are:
- no usage of AI to commit murder WITHOUT a human in the loop
- no usage of AI for domestic mass surveillance
So... this would be fine with them?
Claude: "Are you sure you want me to commit murder?"
User: "Yes"
Or do you mean Human presses button:
Claude: "Do you to commit murder? If so press the button."
User: "I pressed the button"
Claude: "Great! Now lets summarize what we did."
First one
Seems like an absurd distinction to me... Reminds me of "I was just following orders"...
I mean the distinction doesn't really matter
There are many ways to construct HITL UXes. But typically they'd take the form of the first one
I think you're missing the forest for the trees. All Anthropic is saying is that HITL is required before murder, the UX is irrelevant
What i'm saying is yes the UX is completely irrelevant and so is adding a human in the loop before murder.
A real stance would be not allowing your product to contribute to murder of any kind. I think focusing on the petty distinction they are making is missing the forest for the trees.
Thank you for the clarifications though it is helpful context.
Given that Anthropic is clearly risking their entire business just to stand up for what they believe is right, which appears to be what everyone here agrees with, is everyone who is supporting them here planning to also start using Anthropic and switch away from other vendors until they follow suit? Or are folks planning to just use whatever regardless?
Edit: I should perhaps clarify I'm more interested in paid users, rather than free. It's harder to tell if free users switching would help them or hurt them... curious if anyone has thoughts on that too.
My understanding is that they would have been likely to lose many of their senior researchers if they had backed down here.
i’m currently subscribed to openai for their $20 a month tier chatgpt subscription.
i told myself if anthropic does not back down on their current stipulations to the DoD, then i’d cancel and switch over to claude
they said there is a line they do not want to cross, and stuck to that stance, at great personal and financial risk to themselves
I've only ever used the free plans, but I'd consider a sub with Anthropic now.
I'm switching.
This is just an authoritarian state, wanting to use AI to implement something almost certainly anti freedom. We have to be honest about that.
So the government said, We need y’all to flip on the Minority Report and the Terminator modes or we’ll put you out of business… cool
> Our position has never wavered and will never waver: the Department of War must have full, unrestricted access to Anthropic’s models for every LAWFUL purpose in defense of the Republic.
Kesha tried to hug Jerry Seinfeld vibes.
> Anthropic delivered a master class in arrogance and betrayal as well as a textbook case of how not to do business with the United States Government or the Pentagon.
Strange way of saying "this vendor doesn't meet our software requirements".
> they have attempted to strong-arm the United States military into submission
Err... You approached them?
> a cowardly act of corporate virtue-signaling that places Silicon Valley ideology above American lives.
It's an orthogonal point, but "Silicon Valley ideology" has made up a significant portion of the USA's GDP for the last however many years.
> Their true objective is unmistakable: to seize veto power over the operational decisions of the United States military. That is unacceptable.
Again... You approached them?
> I am directing the Department of War to designate Anthropic a Supply-Chain Risk to National Security.
Like most companies in the world I imagine. They just haven't been approached yet.
> to allow for a seamless transition to a better and more patriotic service.
Internally re-framing all the recent "EU moving away from American tech!" articles as "EU builds more patriotic services!"
> This decision is final.
Nothing says "final" like a Tweet. The most uncontroversial and binding mechanism of all communication.
I imagine I'm not the only one to switch over to giving Claude my money today. I'm sure the "Other" comments for the cancellation were often as blunt as mine.
Q: "Is there anything we could do to change your mind?"
A: "Yes! Stand up to the current administration."
Probably used Claude to write the tweet.
"Hey Claude, make this sound less durnk ..."
I'm convinced the only possible good end game here is if this leads to a showdown where GenAI is just made illegal full stop.
Neither side wants that so seems pretty unlikely
In what fantasy world?
A world where I can prompt my local ASI to put a stop to it.
Last I heard, it's still legally called the Department of Defense.
But anyway, I guess the question is, will any other big AI companies stand with them? It's what needs to happen, but I am not hopeful.
This is getting silly guys. All on the same team. Need to have a c.t.j. meeting.
Don't worry, they will be seized by the government soon. Sounds crazy right. Not that far from the headline though, that would sound insane a mere 18 months ago.
Insanely stupid and petty decision. I just left voicemails for all my members of Congress urging them to fight back. I hope the DoW loses this one.
https://xcancel.com/secwar/status/2027507717469049070
The whole thing is fascinating. In my heart of heart, in principle, I want models to be essentially unrestricted, but I still find it somewhat problematic that government thinks it can say: you will make adjust your product to match our exact expectations even if you don't sign an updated contract with us. Odd stuff. I know they are trotting out War powers, but.. well.. we are not at war ( at least not yet or at least not yet officially declared.. ).
Oh well, I guess I've got no choice but to sign my business up for Pro plans with Kimi K2.5. lol.
Google and Amazon both partner with them and sell to the US Government... so does this mean they can't run on Google or AWS infrastructure?
Working with the government is typically a huge pain in the ass unless you have a lot of friends on the inside. It's not hard to do the math when you you dealing with a government whose acting incredibly oppositional.
I had the co-founder of Levels and current head of the US Treasury Sam Corcos reach out to me a few weeks ago for a job. I was initially kind of excited because I had really wanted to work for the Treasury a couple years ago, so I took the phone call with him.
He called me and he seemed like a nice enough guy, but I realized that he's one of the DOGE/Elon acolytes and he started talking about how he's "fixing" the Treasury and that every engineer is apparently supposed to use Claude for everything.
It would have been a considerable pay downgrade which wouldn't necessarily be a dealbreaker but being managed by DOGE would be, but mostly relevant is that I found it kind of horrifying that we're basically trusting the entire world's bank to be "fixed" with Claude Code. It's one thing when your ad platform or something is broken, but if Claude fucks something up in the Treasury that could literally start a war. We're going to "fix" all the code with a bunch of mediocre code that literally no one on earth actually understands and that realistically no one is auditing [1].
If they're going to "fix" all the Treasury code with stuff generated by Claude, I'm not sure they will have a choice but to stick with it, because very it seems very likely to me that it will be incomprehensible to anything but Claude.
[1] Be honest, a lot of AI generated code is not actually being reviewed by humans; I suspect that a lot of the AI code that's being merged is still basically being rubber-stamped.
don't worry
it won't be the world's bank for very long
There's an awful lot of momentum with the USD being the world currency. Even if it eventually declines I think it might take decades, if the British pound is anything to go by.
the UK hadn't fucked off every single one of its allies in the space of 12 months
Trump will default on the national debt before the end of his term.
I am fine with this. If you are a defense contractor, you are a defense contractor, and you follow the military needs that you government believes are necessary - or you stop being a defense contractor.
I wouldn't want a bullet manufacturer to hold back on my government based on their own internal sense of ethics (whether I agreed with it or not, it's not their place)
You're fine with a company being designated a supply chain risk, a designation heretofore used exclusively for foreign adversaries and usually a death knell for most companies, because the government wants to break a negotiated terms of service and contract that they already accepted?
The fuck?
What does this mean for Bun (recently acquired by Anthropic)?
Confirmed: we're living in hell.
Maybe time for Anthropic to leave the US. Come to Australia :)
Under normal circumstances this would end up in court, but when this administration ignores court orders it doesnt like Anthropic would effectively have no legal recourse.
What court orders has the admin ignored?
Here's one specific case[0] and an article citing 35 others over the course of 6 months[1]:
[0]https://www.ap.org/news-highlights/spotlights/2025/unquestio...
[1]https://www.nytimes.com/2026/02/23/us/politics/judges-contem...
The next question, what person wants to send all their personal questions to whichever AI lab does help the government do domestic surveillance
- Co-authored by Claude
Since google aws have contracts with the governor, can they make cloud providers stop providing services to anthropic?
https://x.com/PalmerLuckey/status/2027500334999081294
It is an interesting point. What's the difference between this use license and others?
What a dork.
If the government thinks the terms of Anthropic are unacceptable, they can just stop using them, right? But why would you then retaliate and ban other companies from making business with Anthropic if they want to be a defense contractor? How do these requirements make Anthropic a supply chain risk that makes them unusable for use by other companies?
> If the government thinks the terms of Anthropic are unacceptable, they can just stop using them, right
That is what they are doing.
> why would you then [....] ban other companies from making business with Anthropic if they want to be a defense contractor
Because, if it shops with Anthropic code, the DoD becomes subject to the restrictions when they receive the contractor's product. Anthropic's limitation is on the use, not on the product or distribution.
To stop using them requires making the suppliers still using them as well.
It's perfectly reasonable for the US government to end the contract if they no longer like the terms they agreed to (assuming the contract does in fact let them); it's not reasonable to destroy the counterparty to the contract in retaliation. The line "I am altering the deal; pray I don't alter it further" is literally spoken by Darth Vader, the most comic-book of comic-book villains.
Then the government should end their contract with Anthropic. The terms of the contract were clear.
Designating them a supply chain risk is unprecedented authoritarian strong-arming.
What player is going to step in and do what Anthropic wouldn't? Or, worse, will the DoW try to author its own AI to go where private AI won't?
Probably Grog, which probably means even worse outcomes
At least we'll have hyper sexualized child soldiers to look forward to in our upcoming xAI powered civil war!
Grok is already being brought in
> Anthropic’s stance is fundamentally incompatible with American principles.
I don't think that Secretary Hegseth is qualified to speak on American principles.
Cheating on multiple spouses[1], being an active alcoholic, and being accused of multiple sexual assaults and paying off the accusers[3] is fundamentally incompatible with being a Secretary of Defense and a good leader.
Also, this violates freedom of speech and will probably get shot down in the courts.
1. https://en.wikipedia.org/wiki/Pete_Hegseth#Marriages
2. https://en.wikipedia.org/wiki/Pete_Hegseth plus multiple recent media pieces
3. https://en.wikipedia.org/wiki/Pete_Hegseth#Abuse_and_sexual_...
I got downvoted for this in the other thread, but this is basically an attempt at bankrupting Anthropic. No US company has ever been designated a supply chain risk, and the foreign companies that are on that list are now doing 0 business in the US. Very large portion of the US economy relies on some contracts with the US government, Anthropic cannot survive this if this holds.
I don't think it will hold, in the end this is mafia behavior, but if it does, we are yet again in uncharted waters.
I'd at least, you know, pretend we had a top-secret amazing model. By airing all of this publicly, they've basically admitted that Claude is the best there is.
Besides just being yet another example of the Trump admin abusing power and weaponizing legitimate laws in illegitimate ways to extract concessions, there is another reason this is dumb -- which is that Anthropic just has the best models!
As someone who wants America to win, ripping out Claude and putting in xAI is a terrible idea. Definitely setting us back a few months on capabilities
Wild that not wanting to support fully autonomous weaponry…yet…is the sane take here.
Bluster followed by a "we can't do it now but we will... soon". Whoever has the best model can do what they please you'll see. I work with these things daily as an engineer (been doing this shit for 25 years and wow it's like mana from heaven these days). Believe me no one is going to screw with themselves by not using the best one and right now Anthropic has it.
Anthropic’s stance is fundamentally incompatible with American principles.
Come to EU guys, we'll prepare a warm welcome!
EU won't do 996
Not doing 996 is a feature not a bug
Not when you want to win and compete with someone who does 996
Anyone who does 996 is being exploited, unless they're the actual boss, in which case they're the ones doing the exploiting if they're pushing 996 on their employees.
This is why 996 bosses think AI can replace their employees, because they already see the employees as robots, not humans.
instead of running guys to the ground, you _could_ hire more people and do shifts if it's that important to stay current.
No, it's 996 for 845 wages.
Anthropic would disagree seemingly.[1]
[1] https://www.businessinsider.com/996-work-culture-silicon-val...
We have other places outside of France, come on!
As in live a healthy life so you can make your work hours more productive?
> Anthropic’s stance is fundamentally incompatible with American principles.
TIL Fully automated killbots and mass domestic surveillance are American principles.
I mean, I should have known but there's no clearer sign saying "leave the country now if you don't agree with this admin" than now I guess.
Its one thing to say "we cannot abide by these terms, so let's part ways", and its another entirely to respond this drastically. The Trump administration will look back on this decision as the most consequential in their efforts to win the 2026 midterms and Republican efforts in 2028. This is a $400B+ American company that has significant partial ownership from Amazon, Google, and other private equity sources; they just made serious enemies in SV, many of whom supported Trump in his 2024 election victory.
This is a pimple on the arse of said consequence. It's one tiny thing in a chain of many bigger things.
It's magnified because it's right now, but this won't affect midterm results barely a whisker compared to many other daily headlines.
There are no serious enemies to this administration in SV and I can't see this changing that. SV has bent the knee exactly like Anthropic didn't. They're not going to stand up because of this, they've proven they don't have those muscles.
OTOH it could amplify their base: “Big Tech refusing to work with us on National Security matters!” The base will never hear what/where the red line was drawn, just that Some Company in California (liberal/bad) is being Woke and Political.
Good, anthropic should sell there services to China introduce the “security risk” to China.
The funny thing about stupid people, they do stupid things all the time...
Will be interesting to see how quickly it becomes clear that most of Anthropic's competitors are stealing from them.
I can't seem to find what being designated a "Supply-Chain Risk to National Security" implies from a legal standpoint. From what I can find, it doesn't seem to be a formal legal status. Curious if anyone knows more.
Basically, if you are a federal contractor, the designation means the DoD can force you to certify that Anthropic tech is not used in the fulfillment of your government work. Because it's just a DoD designation, and an executive order and not added to the NDAA, you can still use Claude for non-government (federal) touching work.
So using Claude Code to write software for the DoD is now a no go, you'd be in breach of procurement directives now.
If they go as far as to convince congress to add Anthropic to the NDAA, that would be a nationwide ban like Huawei making it illegal for any federal contractor to use the tech anywhere in their business.
But for now, even fed contractors can still use Claude in their business, just not directly for government work.
That doesn’t seem to match up with the original tweet though - it sounds a heck of a lot stronger:
> Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic
Emphasis mine.
And I’m looking at news organizations that presumably have staffs of legal analysts pouring over this stuff, and they also seem to be saying that it can’t be any commercial activity:
> The label means that no contractor or supplier that works with the military can do business with Anthropic.
https://www.nytimes.com/2026/02/27/us/politics/anthropic-mil...
We have a terrible government. I think that’s the answer.
This is the most unhinged thing yet, after all the previous unhinged things.
Hey Anthropic, Europe welcome you!
Unserious people, in the most serious of positions.
What's with the Republicans. Do they want a strong or a weak government? I can't tell anymore.
I don't think it's ever been about strong or weak, or at least I don't think that's where the differentiation is. You always want 'strong' government, committed to the things it says it's committed to.
It's more been about the size of the government; that it should do a minimal amount of control (and do it well), but leave a lot of things for "the market to decide".
Having said all that, I think this issue is just tangential to any big/small government ideology. This is a hissy fit about a defence contractor sticking to their agreement where the DoD want to change the agreement in a way that goes against the contractors Mission Statement and/or the US Constitution itself.
The old ideology of the Republicans doesn't mean anything here. This administration is purely about 'give me what I want, now!'.
And it's whims change with the breeze. Do not look for consistency here.
It'll get cleared up.
TACO
Look folks when he's (trump) that stuck on stupid, he's right and you're wrong. Class it up, people! Class it up!
The 20th century is finally over...
Something is clearly unraveling.
The US Government is such a bunch of clowns - it's hard to take their nonsense seriously... well except that their stupid policies kill people...
Why does this feel like a Facebook post from the person who got broken up with
Wonder what other countries are doing in this situation
This whole tweet seems very childish.
How 'bout that government meddling in the free market, eh?
Every conservative needs to do some very deep, very serious soul-searching. As for me, as a hyper-progressive, I'm drawing up proposals for nationalizing real estate developers in order to force them to build new houses to sell below cost.
In all this commotion I've completely forgotten that Anthropic dropped their safety pledge three days ago.
Finally silicon valley is being shown who they sucked up to.
Can we all take a big step back and just ask why the DoD wants to use a fundamentally unreliable technology to guide deadly weapons?
The same reason why they used a Signal chat group for discussing matters of national security.
They don't. They want to punish a company for expressing values that introduce friction to the whims of the current administration.
No, stop, I understand the politics here, but I’m asking about the technical fundamentals.
LLMs produce output of unknowable and unpredictable accuracy, and as far as we know, this is a mathematically unsolvable problem. This shit should not be within 1000 miles of a weapons system. Why are we even talking about this?
> LLMs produce output of unknowable and unpredictable accuracy
So do humans. But humans might not follow illegal or immoral orders.
The DoD killing lots of people based on faulty intelligence - never!
Joking aside, this administration clearly cares much less others. They don't care if innocent people are killed.
Because of the politics.
I am directing my Department of Peace to designate Anthropic as a Supply-Chain Risk to Fascism.
I have just purchased a chunk of extra usage credit. I encourage my peers to do the same. Let's send a message to those that work forces.
So the DOW is using it till the mid term elections?
>Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic.
Nevermind Claude, does that mean Anthropic's offices can't use a power company if that same company happens to supply electricity to a US military base? What about the water, garbage disposal, janitorial services? Fedex? Credit card payments? Insurance companies? Law firms? All the normal boring stuff Anthropic needs that any other business needs.
This is a corporate death penalty. Or corporate internal exile or something, I don't know of a good analogy.
A level up, this is only the beginning of the political headwinds for AI. There will be a lot more, especially if constituencies begin to get displaced. I don’t think “job loss” will really occur, at least not in a dramatic way overnight. But I do believe there will be both aggressive regulation and very aggressive taxation of this technology in the near/mid-term.
we are experiencing marketing at its best
They should wear it like a badge of honor
Good. At least now I don't have to worry that my vibe-coded, unreviewed checkout button is accidentally going to hallucinate the command that blows up a kindergarten in Yemen.
We can actually get a glimpse of how AI might wipe out humanity here.
Model collapse making models identify everyone as a potential threat who needs to be eliminated.
Companies should have a right to refuse such requests on moral grounds though.
This stance is vindictive. Just don't use Claude in the military. Extending it to all government agencies is not right. They do great work. Can't deny that.
The US is such a shit show. Personally I hope this doesn't affect Anthropic's growth and development because I quite enjoy using their products and see them evolve.
i think this is just a show they are putting out .
This is only the first year of this fascist government, and I believe the first powerful company that is taking a stance? Meta, Apple, etc. have all bent the knee right?
Apple not just bent the knee, but also presented a golden plaque to go along with it. Yuck
Stop calling it the Department of War, it's not the official name of that agency.
Department of War is a teenage boy's idea of "manly" and "cool". Same with X. These juvenile idiocrats will be laughed at by children in the future studying history. "Seriously? How dumb were these people in the 21st century."
Grok in US gov in 3 2 1…
Already there 'February 23, 2026: The Pentagon confirmed a new agreement allowing Grok use in classified systems. Defense Secretary Pete Hegseth announced it would go live soon on unclassified and classified networks, alongside other models, as part of feeding military data into AI.'
This will mean Grok becomes the defacto US Gov AI provider.
Stupid situation, but a badge of honour awarded to Anthropic.
Why are so many adopting this name for what is by law, by the American people, called the Department of Defense? The name change pertains directly to the Anthropic issue, which is the function of the government and department, the power of the American people to govern themselves, and the role of the president relative to the soveriegn American people.
Well put and it bothers me too. It seems to be another case of Orwellian manipulation, i.e. an expression of power through language, functioning as a litmus test of the speaker's loyalty. Serious publications are not going along with it. More craven or (here) thoughtless ones are falling in line.
I mean the original name switch was much more "Orwellian manipulation" if anything changing it back to war is undoing the bullshit implications that everything it does is defense.
Surely the purpose of the organization is to defend the country? War seems more like the failure mode. The point here is that it was established by a law of Congress and so has an official name that should be respected until another law changes it.
Because it sounds a lot cooler.
OpenAI came out just last night or today claiming they would hold the same line as Anthropic. Makes me think both sides knew Elon had already won the contract.
Cue xAI.
And here’s the irony: Musk, who claimed only he is virtuous enough to defend us from AI, who insisted he always wanted model labs to be non profit and research focused, will now bring his for profit commercial entity into service to aid in mass domestic censorship and fully autonomous weapons of war.
In fact it won’t surprise me further if NVIDIA is strong armed into providing preference to xAI, in the interest of security, or if the government directly funds capital investments.
Anthropic saves some dignify and they’re the losers today, but we are the losers tomorrow.
I can't wait to read the transcript of the AUSA in front of a federal judge trying to explain threatening to declare a company a supply chain risk if the company doesn't supply things to the government.
> Anthropic's two hard lines:
> 1. No mass domestic surveillance of Americans
> 2. No fully autonomous weapons (kill decisions without a human in the loop)
Surveillance takes place with or without Anthropic, so depriving DoW of Anthropic models doesn't accomplish much (although it does annoy Hegseth).
The models currently used in kill decisions are probably primitive image recognition (using neural nets). Consider a drone circling an area distinguishing civilians from soldiers (by looking for presence of rifles/rpgs).
New AI models can improve identification, thus reducing false positives and increasing the number of actual adversaries targeted. Even though it sounds bad, it could have good outcomes.
I thought Anthropic's take on #2 was they don't think the model's good enough yet?
But compared to what - if Anthropic's models aren't perfect but still better than existing (old school) models, it's understandable DoW still wants to use them (since they're potentially the best available, despite imperfections). I think Hegseth is saying to Anthropic: "that's our call, not yours".
But surely if Anthropic thinks there's a risk that their models might make bad decisions, and the resulting civilian or etc deaths are blamed on them, it's their right to refuse to sell it for that purpose? That's why they had those restrictions in the contract to begin with. How can they be forced to provide something?
I agree they can't be forced to provide something. I just see DoW's reasoning, and I can't fault it.
Anthropic are taking a moral position which is admirable, but in this case it could actually make people's lives worse (if we assume more false positives and fewer true positives, which is probably a fair assumption given how much better 'modern' AI is compared to the neural net image recognition of just a few years ago).
based
Such a dipshit administration. I hope California secedes from the union to protect our champions.
Batshit situation, respectable position from Dario throughout.
But there's some irony in this happening to Anthropic after all the constant hawkish fearmongering about the evil Chinese (and open source AI sentiment too).
Fuck it, I am buying a Max Pro subscription just because of this.
"I am altering the deal. Pray, I do not alter it further." - a scary evil dude.
Theo's got a good overview
https://youtu.be/MWFyApldYDA?si=yskCcx2hY4Wjkgw8
Please tell me when their fifteen minutes is over. It is one bad joke after another.
An earlier post to a news article rather than to a tweet: https://news.ycombinator.com/item?id=47186662
That news article doesn't mention the designation of Anthropic as a supply chain risk (it was published about 20 minutes before Hegseth's tweet)
Hegseth's had a busy week: trying to kill Anthropic, attending the State of the Union, fighting Scouting America, and his regularly scheduled efforts to shame fatties & trans kids... Unlike so many in the orange one's inner circle who are just incompetent (say, Kash Patel for one), this dude is both incompentent a very bad, bad person.
I am reminded of bcantrill's legendary quote:
> You don’t anthropomorphize your lawnmower, the lawnmower just mows the lawn - you stick your hand in there and it’ll chop it off, the end.
Except this is like two lawnmowers going at it, which would be a sight to behold indeed.
AI crash here we come
when do they go to court?
I've had issues with Anthropic since the beginning. I never trusted them. Whoever did, might have some problems.
Can we get a list of companies with this designation so I can migrate my subscriptions to them?
let's see...
> Populist nationalism + “infallible” redemptive leader cult
> Scapegoated “enemies”; imprison/murder opposition/minority leaders
> Supremacy of military / paramilitarism; glorify violence as redemptive
> Obsession with national security / nation under attack
TBH could be worse.
Trump's associated "Truth" ("Truth Social" is the name of his risible fake-Twitter and they call Tweets, "Truths" there) that preceded this:
https://www.trumpstruth.org/statuses/36981
Don't worry, this is an archive/mirroring site for his account, not the actual TS site.
I'd comment on how wackadoo this all is, but, 1) that applies to almost everything these days, and 2) the post's right there, see for yourself.
I really don't follow USA-politics besides the occasional hn-thread, random yt videos, and comments from friends...
With that said: what are the chances, in your opinion, that Donald wrote that himself?
To me it reads too coherent for there to be any chance he wrote or even dictated that.
I think odds are high a lot of these posts are by staffers. The posting volume is bananas, even granting that he spends a lot more time personally online and watching cable news et c. than any prior president, I don’t think there’s any way they’re all by him.
I do think a lot of the more hot-take type posts (often in response to stuff he’s watching on tv) are either actually him, or he’s dictating to an aide. These larger policy-type ones that he treats as quasi-executive-orders, I think are likely drafted by one or more of his cabinet-level folks, or others roughly as high up. That’s just my speculation based on reading the “tea leaves”, though.
As for official word, it waffles between “all of it’s him” and “oh not that one though, that racist video repost was a staffer who made a mistake”, so that’s little help in sussing out the truth (but I am rather certain they’re not all directly written and posted by him)
Sigh. So dumb.
More taxpayer funded lawsuits to come.
This is going to have two unintended consequences.
One, it’s going to fuck with the AI fundraising market. That includes for IPO. If Trump can do this to Anthropic, a Dem President will do it to xAI. We have no idea where the contagion stops.
Two, Anthropic will win in the long run. In corporate America. Overseas. And with consumers. And, I suspect, with investors.
> In corporate America
A lot of corporate America contracts for the military in some capacity (it's a giant piggy bank and if you jump through a few hoops you get to siphon money out of it, so of course they do) and assuming this Tweet is accurate (Jesus, what a world) this will also affect them.
IDK maybe they have corporate structures that avoid letting this kind of thing mess too badly with the parts of their company that don't have contact with the government, or maybe it'll only apply to specifically the work they do for the government, but otherwise I expect it'll be devastating for Anthropic's B2B effort.
> lot of corporate America contracts for the military in some capacity
And a lot does not, or does so through dedicated subsidiaries so they can work multinationally.
What percentage of their revenue comes from the government?
Pete Kegseth is unhinged. I’m siding with Anthropic here
AI proponents have been very vocal about AI safety being meaningless. But nobody could have expected that the end of the world would have come because Trump puts Grok in charge of the US nuclear arsenal. We truly live in the dumbest timeline.
Related:
Trump orders federal agencies to stop using Anthropic AI tech 'immediately'
https://news.ycombinator.com/item?id=47185528
Statement from Dario Amodei on our discussions with the Department of War
https://news.ycombinator.com/item?id=47173121
And the White House, quoting Donald Trump: https://xcancel.com/WhiteHouse/status/2027497719678255148
"THE UNITED STATES OF AMERICA WILL NEVER ALLOW A RADICAL LEFT, WOKE COMPANY TO DICTATE HOW OUR GREAT MILITARY FIGHTS AND WINS WARS! That decision belongs to YOUR COMMANDER-IN-CHIEF, and the tremendous leaders I appoint to run our Military.
The Leftwing nut jobs at Anthropic have made a DISASTROUS MISTAKE..." - President Donald J. Trump
No surprise here. All government actions are now in the Trump mafia boss style.
“You won’t let us use your product unrestricted for military applications? Fuck you, we’re going to stop using it for anything at all across the entire federal government, even if not remotely related to military.”
Kudos to anthropic for standing up for their principles. Let's remember all the silicon valley leaders who embraced fascism without even needing to be pressured. We need more billionaires with backbones.
Hey Hegseth ...
....................../´¯/)
....................,/¯../
.................../..../
............./´¯/'...'/´¯¯`·¸
........../'/.../..../......./¨¯\
........('(...´...´.... ¯~/'...')
.........\.................'...../
..........''...\.......... _.·´
............\..............(
..............\.............\...
When people like you post stuff like this, I know my vote for Trump was the right one.
This makes no sense. Do you vote based on principles and policy or do you vote based on the behavior of people who have nothing to do with government?
I don't think I'll ever be able to understand how someone can read what Trump posts and think "Yeah, that's a guy I want as my President."
ah yes, fascism
Cancel culture and derangement syndrome. This admin is garbage.
Hegseth gets so belligerent when he's hammered.
As best I can tell, his hard-drinking era ended many years before he entered the cabinet. But this does feel like a pretty impulsive decision, and there's some ambiguity over whether this statement was approved by the WH, or whether this was just the SECDEF taking it to the next level to look super loyal and badass. This ambiguity gives the WH room to walk it back in the coming weeks, depending on how things evolve.
I can honestly understand both positions. The U.S. military must be able to use technology as it sees fit; it cannot allow private companies to control the use of military equipment. Anthropic must prevent a future where AIs make autonomous life and death decisions without humans in the loop. Living in that future is completely untenable.
What I don’t understand is why the two parties couldn’t reach agreement. Surely autonomous murderous robots is something U.S. government has interest in preventing.
> it cannot allow private companies to control the use of military equipment.
The big difference here is that Claude is not military equipment. It's a public, general purpose model. The terms of use/service were part of the contract with the DoD. The DoD is trying to forcibly alter the deal, and Anthropic is 100% in the clear to say "no, a contract is a contract, suck it up buttercup."
We aren't talking about Lockheed here making an F-35 and then telling the DoD "oh, but you can't use our very obvious weapon to kill people."
> Surely autonomous murderous robots is something U.S. government has interest in preventing
After this fiasco, obviously not. It's quite clear the DoD most definitely wants autonomous murder robots, and also wants mass domestic surveillance.
So what your saying is it should be removed from the military supply chain?
Because the current government wants unquestioning obedience, not a discussion (assuming they were capable of that level of nuanced thought in the first place). The position of this government is "just do what I say or I will hit you with the first stick that comes to hand".
A vendor doesn't want to do something you need, you find another vendor (there are others).
This is just petty.
If the government doesn't want to sign a deal on Anthropic's terms, they can just not sign the deal. Abusing their powers to try to kill Anthropic's ability to do business with other companies is 10000% bullshit.
> What I don’t understand is why the two parties couldn’t reach agreement. Surely autonomous murderous robots is something U.S. government has interest in preventing.
Consider the government. It’s Hegseth making this decision, and he considers the US military’s adherence to law to be a risk to his plans.
I can see both sides as pertains to Trump's initial decision to stop working with Claude, but now, this over-the-top "supply chain risk" designation from Hegseth is something else. It's hard to square it with any real principle that I've seen the admin articulate.
> What I don’t understand is why the two parties couldn’t reach agreement.
Someday we'll have to elect a POTUS who is known for his negotiation and dealmaking skills.