I know everyone says this is political but it actually seems like a textbook designation. Anthropic wanted to have rules on how the military used AI, the military said no and therefore doesn't want anthropic used anywhere in their supply line.
This is like a pen manufacturer not wanting their pens used to sign drone strike orders, now the military needs to have a special box of pens that don't have stipulations attached. With AI usage it would be the same thing except applied to entire product chains. It seems like it would just add more complexity to operations.
You can agree with the rules anthropic wanted, but having rules set by a private company at all that apply to the military does seem fair for the military to object to.
>The Department reasonably feared that Anthropic might manipulate Claude’s design to prevent it from performing national-security functions that the Department deems contractually authorized and necessary
Though they'd probably put the DoD on the cybersecurity whitelist today, the very idea of the claude whitelists for certain functionality already exists and is being used by them today.
> I know everyone says this is political but it actually seems like a textbook designation
It literally is a textbook definition, signed into US law:
“Supply chain risk,” means the risk that an adversary may sabotage, maliciously introduce unwanted function, or otherwise subvert the design, integrity, manufacturing, production, distribution, installation, operation, or maintenance of a covered system so as to surveil, deny, disrupt, or otherwise degrade the function, use, or operation of such system (see 10 U.S.C. 3252).
To add onto what another commenter said, the pen analogy would be more like the manufacturer designing pens that stopped working when used to sign strike orders they disagreed with.
Selling a pen labeled "this pen will refuse to sign certain orders" is not sabotage or malicious and is thus not a supply chain risk. The DoD is free to not buy from Anthropic, but designating them as a supply chain risk is incorrect, as well as arguably arbitrary and capricious given their public criticism of Anthropic's beliefs.
I don't think anyone doubts that subcontracts exist. They doubt that Anthropic's insistence that Claude isn't capable of being the operating system for an automated killbot and associated terms the DoD previously agreed to means that there's any danger to the US military from Lockheed Martin using Claude as a code assistant to build GUIs.
What if it’s a GUI to manage deployment of the kill bots? I worked for a DOD contractor on software. All our scenarios were more related. It was like “OK you’ve overthrown the civilian government of this Middle Eastern country, including the radio regulators, and now we need to coordinate radio frequency assignments for the tactical squads.”
Code which renders dots on a map isn't going to magically reorder itself if the GNSS endpoints get switched from manned to unmanned assets just because it was authored by Claude, and if the subcontractor isn't auditing and testing the code outputs properly then they're the supply chain risk, not a specific code-gen tool they use. Obviously if instead of being used as code or report generating assistants, LLM agents are used to execute combat decisions in real time that introduces a lot more risk. That's... literally Anthropic's argument against using them for that specific purpose.
It doesn't make much sense to determine that because Anthropic continues to insist that Claude is unsuitable for use as some sort of super advanced loitering munition, Lockheed using it to write job ads is as much a threat to national security as a Chinese comms terminal.
It's not like the DoD and their subcontractors and their subcontractors' subcontractors aren't used to all sorts of fine-grained separation when it comes to what is and isn't classified or export controlled under normal circumstances, or that Hegseth taking a break from fantasising about soldiers' testosterone levels for long enough to decry Anthropic's stance as 'woke' makes the decision look like one based on operational effectiveness.
"Supply chain risk" specifically refers to risk of malicious attack or sabotage through the supply chain, not all risks associated with supply chains.
A vendor with particularly poor QA might be a risk in your supply chain but isn't a "supply chain risk" according to this designation and neither would Anthropic be.
Supply chain risks are usually when someone does a attack over the supply chain. Like when the mossad planted literally explosives in radio devices for Hezbollah.
Antrophic merely said they don't allow their tools to be used for autonomous killing.
A very specific case.
So where is the danger making it necessary preventing all suppliers from using Claude at all? That only concerns those who work directly on autonomous killing and only they will have to go to Altman or Musk.
So none of that is any threat to the military - so it was just black mailing to make them do the governments will - which is something usually dictatorships do, not democratic systems.
In a democratic society, the default should be, the military just buy somewhere else, then force companies to build them the terminator. So no private company controls what the military does, they just control their product. You want the military to control private companies? There is a name for that kind of system, where this is conmon.
Eh this just we wont buy from you because you are putting limitations on the use. Anthropic will be fine if they don’t like it they can accept the terms of the contract or walk away.
Why has this never previously been done to an American company from which the DoD decided not to purchase something?
Perhaps people who are noticing politics here - including in the decision of two Trump-appointed judges, against the he vote of another Republican-appointed judge- are not the ones who are choosing not to see reality?
President Truman literally seized the assets of a steel company to avert a strike that would hurt the Korean War effort. The Supreme Court stopped him, but three Justices voted it was okay!
We also have the Defense Production Act: https://www.congress.gov/crs-product/R43767. It doesn’t happen more often because most people aren’t stupid enough to say “no” to the U.S. military. These Effective Altruists, however, are too big for their britches.
I don't know, I am just guessing because not a single American company has publicly stated they are against its product use up to a certain degree by DoD? If they have not publicly declared it then yes, DoD simply don't sign the contract.
So your argument is that the government is punishing Anthropic for its public statements? And it is correct to do so?
(I think it's also the case that Anthropic didn't make any public statements about the negotiations until after the DoD was already publicly threatening this designation. But that seems less important than the question of: is it good for the US government to try to destroy companies whose public statements it dislikes?)
Honestly I would find a certain satisfaction in a world where the next Democratic administration takes away Fox broadcast licenses, forbids use of Oracle by any government contractor unless Ellison hands CBS over to Rachel Maddow, and tells Bezos that Amazon is going to be banned from public roads unless the Washington Post front page has positive stories every day. But I doubt the same is true for the many pro-Trump people who complained vociferously about Biden officials expressing preferences to social media companies because of the merely implicit threat that they might take government action of some kind if they weren't listened to.
So we have a pen that may refuse to sign certain orders, and is clearly labeled as such.
If the DoD needs to sign such orders, they won't buy this pen.
If a supplier needs to sign such orders, the supplier also won't buy the pen.
If a supplier needs to sign other orders, and this pen is the best one available, the supplier may decide to buy this pen. Since the orders are within the range of what the pen will sign, they get signed, and the supplier fulfills its contract with the DoD.
Where is the supply chain risk? The ink isn't going to erase itself post-hoc.
---
Now, if the DoD suspects that Anthropic will train its models to try to actively sabotage military operations, that's a very different story. Does anyone actually believe that?
> If a supplier needs to sign other orders, and this pen is the best one available, the supplier may decide to buy this pen. Since the orders are within the range of what the pen will sign
Who decides what’s within the scope of “what orders are within the range of what the pen will sign?” The boundaries are never clear cut. You’ve got a vendor who could kill switch critical military technology if it’s used in a way they decide falls outside the scope of what they want.
> You’ve got a vendor who could kill switch critical military technology if it’s used in a way they decide falls outside the scope of what they want.
If the DoD hires ACME to build some software to make widgets, and ACME uses Claude to build that software, where is the kill switch?
I guess the Claude model could hide a kill switch in the widget manufacturing code, but so could ACME's human subcontractor. Why is Anthropic being considered a supply chain risk here?
The model can refuse to build something, but it can't take back what has already been built.
If you outsource your thinking to a server that can detect your IP address, and decide to render decision “b” instead of decision “a” to your query, that’s a risk. (Or detect by the type of query, past history of your use, etc.). Panopticlick should be having a field day with all this.
Oddly when the topic of e.g. car infotainment systems comes up, like BMW locking you out of features in your own car, this place is up in arms, but when the military has a problem with it, suddenly they should be forced to accept it?
All of those verbs basically mean "degrade an existing system, after the fact". If anthropic states beforehand that certain use cases are not possible, it should be fine, no?
To give another (bad) analogy: The Pens aren't made to be weapons themselves, they are not intended for that. If the DoD now tries to shoot pens as projectiles and they just melt, that's no reason to classify the pen manufacturer as a supply chain risk, the product was not altered after the fact.
Of course, there _is_ a difference in "It could do that, but it shouldn't" and "It cannot do that". But it is not that large. That's why I don't see it as that clear cut.
The pen example makes sense if the intermediary were to ship the pen as part of the final product but not if the pen is simply used to draft designs of the final product. It seems that this designation would prohibit using the pen anywhere in the process which doesn't really make sense.
> The U.S. military has disabled advertising tracking tools in government-issued phones, computers and other devices, officials recently told Congress.
> The different military branches confirmed they had done so, some only this summer, following reporting that commercially available data taken from these devices and sold freely by data brokers, was being used by Iran and other adversaries to track and in some cases target American bases and personnel.
the courts interpret the statute, we have to read their argument where they explain why they think that's not the right definition to focus on. (the "adversary" is in section 3252, but they say that independently from that section 4713 allows determination of exigency.)
Ah, I upvoted specifically because those countries do not control Anthropic's products, and thus reasoned that the comment implied that an adversary, for the purpose of the law, was some external not-quite-enemy type entity, and that application of the law to Anthropic was thus unreasonable, as it obviously is.
Of course they don't but the post was responding to "what adversaries" in the definition of supply chain risk posted above, and those are some of the real answers to that. I did not read that comment as implying anthropic was controlled by those countries at all, just responding to the actual question. My take away from the post was that those are actual adversaries and, unless the administration is claiming anthropic products ARE controlled by those adversaries, the claim of "supply chain risk" is not in good faith.
While amusing on the surface, let's substitute an ink-jet printer for the pen, and have a malicious party inject code as a print driver update, or maybe re-programs that chip in the ink cartridge, and the ink cartridge refuses to function on certain print jobs. Maybe it will only allow the words "WE SURRENDER" be printed as part of the message.
Here on planet Earth, what happened is that Anthropic imposed licensing terms. The DoD is free to reject those terms. But that is not evidence that Anthropic is planning to harm the US military by sabotaging its products. There is no evidence that Anthropic changed, or planned to change, the design of its system to work against the US military following its dispute with the DoD. Zip zero nada. Never happened. Anyone who tells you otherwise is lying.
>To add onto what another commenter said, the pen analogy would be more like the manufacturer designing pens that stopped working when used to sign strike orders they disagreed with.
The pen analogy would be more like the manufacturer selling pens that work just fine under all circumstances but that the military starts using in hand-to-hand combat.
It is not the textbook definition, because what Anthropic is doing is not sabotage, malicious, or subversive. Those are the key words in the definition. They are just refusing to add a feature to the military's specification. So their bid falls short of requirements.
> Whatever paradigmatic examples individual members of Congress may have had in mind, the statutory definition is not limited to “adversar[ies],” 10 U.S.C. § 3252(d)(4), and instead covers “any person,”
I tried to hint in my comment that the argument is completely different. the supply chain risk is just what the media runs with. (yes, it's in the opinion, but that is the weaker part. and the court agrees, that it basically does not apply to Anthropic.)
there's section 4713 which is roughly "Secretary of War can pull the nat.sec. card, bye"
> They are just refusing to add a feature to the military's specification.
That fails to accord with the claim:
>The Department reasonably feared that Anthropic might manipulate Claude’s design to prevent it from performing national-security functions that the Department deems contractually authorized and necessary
That isn't sabotage, that is simply product design. Anthropic is free to create their products that refuse to kill people, and the DoD is free not to buy those products if they don't fit their needs.
However, that's not what a supply chain risk is. It's not an order to "don't buy these products if they don't work the way we want them to", its designating Anthropic as a national security threat because they might intentionally sabotage US military operations.
As I said elsewhere, it's utterly preposterous that the DOD actually considers this a reasonable threat, because it's simply not a reasonable possibility. There's no way Anthropic would do that, precisely because of the consequences that would follow if they did, and got found out. Moreover, they already clearly stated their terms and preferences. It's all out of the open. There's no supply chain risk, that designation is purely political.
I work in the space industry and regularly use parts in our designs that are unapproved from space flight. Sometimes it's because it's a commercial part that we seem acceptable for flight, other times it's an unscreened or engineering model that doesn't go through the proper testing that space-qualified parts do. One vendor even dents the lid of a part to invalidate the hermetic seal guarantee that they claim for the space-qualified version (it still works fine). Many will have fine print in the data sheets saying the part is not to be used for critical applications like aerospace or medical devices. Sometimes we tell a little fib to our vendor that we are just developing non-flight devices so that they don't get upset and refuse to sell the lower-grade part to us. These vendors are scared of having something fail in an unapproved environment and are unaware of how much risk the customer is tolerating by the use of these unscreened parts.
It would have been incredibly simple for Anthropic to say "we cannot guarantee performance in a kill-chain application due to an unknown level of safeguards implemented in the baseline product and cannot estimate a cost for developing a new product capable of such application." and leave it at that.
>The Department reasonably feared that Anthropic might manipulate Claude’s design to prevent it from performing national-security functions that the Department deems contractually authorized and necessary (emphasis mine)
They didn't "reasonably" fear anything. There is absolutely no evidence to support these allegations. Anthropic has the right to impose licensing terms. Merely disagreeing with the government is not evidence of wrongdoing.
If Anthropic wasn't allowed to say what they said, how would they be allowed to object to their government's illegal use of their products at all? It sounds to me like their only choices were "shut up" or "be punished", which is a clear-cut violation of their rights.
Anthropic, in the original negotiations, said the can't answer every hypothetical use case. And the DOD should just come and ask as they come up (during an operation).
Which means the exact use limitations WOULD NOT BE DELINEATED IN WRITING in advance.
That fails only if you believe and agree with the government's argument, which I don't.
I don't think it's reasonable to fear that Anthropic would change the deal after contractually agreeing to terms of use. The government is using that as an excuse because they know that Anthropic hasn't actually met the definition of a supply-chain risk.
The department feared Anthropic would refuse if it was used to mass surveil on Americans or kill people without human oversight. Facts already laid out in the terms. Instead of simply canceling or negotiating, they used overwhelming government force against it to apply a designation never before applied to an American company. When its an interaction between trump/trump's government and any other party, it is a good first approximation to assume Trumps side are wrong.
Yes certainly, I can sell highly corrosive concentrated hydrochloric acid to the government as part of some contract for the synthesis of certain chemicals but I am fully within my rights to write that the acid should not be sprayed on people from a pump.
But that is not at all a reasonable fear. Is it really reasonable to believe that Anthropic, after receiving a government contract, would then proceed to sabotage their own product to not function as contracted? That seems like an utterly ridiculous claim to me, nothing close to "reasonable". There is no charitable way to view this designation except as political punishment and/or as a favor to Altman and Musk.
> would then proceed to sabotage their own product to not function as contracted?
Claude terms here: ANTHROPIC EXPRESSLY DISCLAIMS ALL IMPLIED WARRANTIES, INCLUDING WARRANTIES OF MERCHANTABILITY, NON-INFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE
Like most so-called AI, the Claude program is inherently unreliable. I doubt Anthropic would ever agree to "function as contracted".
My response was re the not functioning as contracted issue, not the supply chain risk. Having said which, I can well imagine possible cause and effect here.
Yes, but degrading capability on purpose is entirely permitted by a No Warrantee contract, and hence not at all the sabotage their own product to not function as contracted as feared by to DoD. Unless there's a very different and unusual contract in this case.
I don't think you understand what "No warranty" means in a software license. It means Anthropic cannot be held liable for results if the software misbehaves and instead of bombing a terrorist compound it bombs a wedding (I chose this example because the US has a tradition to bomb weddings in Iraq and Afghanistan). But in no form does it mean "yeah, I don't feel like doing what you prompted me".
Hence, Anthropic sabotaged the models on purpose to not respond properly to all prompts.
> It means Anthropic cannot be held liable for results if the software misbehaves
Actually I think it means what is says - which does include no warrantee of FITNESS FOR A PARTICULAR PURPOSE. This lets Anthropic block any purpose it likes.
That's not at all what this is. Anthropic was very clear up-front what they would and wouldn't allow. They were not going to "subvert" anything. "Degrade function" in this context means that there is a risk that ongoing changes to the covered system will cause problems for the government. That wouldn't have been the case here; it would have been clear from the start what the military would and wouldn't be allowed to do with it, and that wouldn't change.
??? What are you talking about. Anthropic isn't an adversary and they're not required to sell for purposes they don't want to. It's not sabotage and it's not malicious.
This is a designation reserved for terrorists and the link.
> If you're preventing the military from doing some action because you disagree with it, that makes you an adversary.
No, it doesn't. The military doesn't have a right to force product requirements and terms of use on companies they contract with.
If they don't like the contract terms, they can simply not sign a contract and not do business with a company. Trying to get them designated a supply-chain risk is just political retribution.
Actively handicapping a product that's otherwise capable qualifies. If you discover that the military is using your capacitors to make rocket launchers, so you introduce a manufacturing defect that will cause them to fail at high heat, you're a supply chain risk and can be rightfully designated as such.
The converse does not qualify. If I advertise a chemical with an admixture that prevents it from being used to make explosives, the military has absolutely no right to demand that I produce a version of that chemical without the admixture. Selling a product which can't do what the DoD wants does not make me a supply chain risk, no matter how easy it would be for me to make a different product that's fit for their purpose.
So... don't buy from Anthropic? How many of the US population is an adversary, according to your definition?
Say everyone had something to sell and DOD wanted to buy it. But some people wouldn't make the product changes DOD demanded. Now, they are the adversary of the DOD?
I think you're confused. When the White House decided to stop having federal agencies buy paper straws [1], they didn't designate paper straws a supply chain risk, they just stopped buying them. The term has a very specific meaning which would not apply in the pen scenario.
It's not a great analagy; it's more like a pen manufacturer not wanting their pens used to sign drone strike orders, who then only ships pens with cameras and solenoids so that the point stays retracted until the camera verifies that the paper doesn't include a drone strike orders.
And that would be absolutely fine for a pen manufacturer to do. It would make them unsuitable for meeting the requirements for a government contract as a pen supplier for signing orders, but would not make them a supply-chain risk.
Same goes for Anthropic: they would only consider offering a version of their product for military use that had certain restrictions. The government is free to accept that version or reject it. Rejecting it and designating Anthropic a supply-chain risk is just political intimidation and retribution for not playing ball.
it would be like the pen manufacturer who ships pens with an automatic point retractor, AND a video livestream to the manufacturer from the pen AND remote control of the point retractor.
anthropic and contractors monitor the livestream remotely to make sure the pen is not signing drone strike orders and may choose to activate the retractor.
> anthropic and contractors monitor the livestream remotely to make sure the pen is not signing drone strike orders and may choose to activate the retractor.
I might be mistaken, but in the case of national security deployments the models are airgapped and Anthropic doesn't actually have access to telemetry/logging from them.
however anthropic is demanding that they have access to chat logs for anything better than opus 5.
it is not possible to run recent anthropic models in a secure way. palantir, nvidia, booz allen etc have blocked fable because they can't run it safely.
You are repeatedly saying they spy on their users? Prove it. Then prove t he other ai companies don't. Then prove that the service they offered to the US government didn't have spying disabled.
"Then prove that the service they offered to the US government didn't have spying disabled."
the service they offered did have spying disabled. the security risk is from anthropic updating the claude and covertly bricking or degrading it.
quote: "In such an ongoing relationship where the vendor [...] “continuously update[s] and tune[s] the product,” the Department must establish and maintain confidence that the vendor will not add limitations
that interfere with the Department’s use of the product." (p26)
"the Department relies
on [petitioner] to provide regular updates to Claude" (p27)
All of these said they keep data 30 days and xai even said furthermore it can decide to keep data for even longer. You also said the version they provided to the government had spying disabled, the complaint that the govt made seems a thin excuse for what trump wanted as any and all major AI models are subject to constant updates.
They're talking like Amazon Bedrock doesn't exist, which is exactly the service which provides private data and processing for models including Claude.
But also they quite clearly have no idea how classified information systems work: air gapping is step 1 of that process, and no one from Anthropic would be working on those systems without a security clearance.
there is no private processing for recent claude models. there is no fable in IL5 or IL6.
bedrock IL5, nvidia, palantir and booz allen have all blocked fable.
anthropic is demanding that they rat out their customers on anything better than opus 5.
see: [thein.fo/4dekujW]
also, bedrock govcloud does both log and review customer data. specifically, they log and review customer data when it is flagged by their csam classifier.
That article is locked behind subscription. These are private companies, and presumably not running the model on their own systems but are using an interface to the model hosted on Anthropics machine I would presume. So their decision to decline is different from the us govt which will have gotten special privileges.
From what I am reading the IL6 and IL7 wasn't granted by the DoD out of nonsensical reasons, it wasn't even allowed to go through the certification process, in fact during the very course of its negotiations for military use of Anthropic. Anthropic never required any remote control or line to the deployment, they only requested there be human oversight before murder. The government prevented it from going through the certification tests using the exact nonsense excuse this case is about culminating in the case. So your claims of spyware are completely muddled and nonsense. 'Spyware' is an extremely absurd and over the top claim. It is a plain and naked lie.
for now it's spyware. there is no way for anyone to get fable without anthropic spying. so there is no fable IL5. there is opus 5 or below.
it is anthropic alone deciding to rat out fable customers.
i am not making this up or lying, everything i claim is verifiable.
anthropic would likely have turned off the spyware for the dod. regardless the dod did not trust them to maintain claude with updates etc.
that is a lot to do with anthropic undermining dod by getting involved with safety at dod contractors (palantir).
going behind the boss and plotting was a bad idea.
it does not help the matter if the bidder rocks up to the pentagon for a $200m contract, with a troupe of bay area ea clowns who say that they don't know if claude is conscious and talk about saving the world from p(doom).
in a battle of ego between trump and amodei the latter is out of their depth. amodei is a newcomer starting a fight with a hardened criminal who has spent years amassing power over the media, the law and the republican party. anthropic are radically outgunned here.
No. You made up the term spyware, the US government stamped a Supply chain risk on anthropic without even letting them get certified, and there is no mention of spyware there. You are making up complete utter nonsense. I have never seen the term spyware used anywhere else in relation to Anthropic. A data retention policy, which your own link shows is also shared with xAI is not spyware. You are a plain liar. You are repeating this absurd lie of Anthropic being 'spyware' everywhere. You again made up nonsense about Palantir, all I can see is Palantir comes under the same case, Anthropic's original stipulations and their objection whether Palantir is following it or not. It has nothing at all regarding spyware. You are trying to blow up one singular point, the DoD's sudden U turn and forcible and false application of a status to appear to be several different unrelated events, and muddling up the cause and effect timeline too. The term spyware is completely absurd for a mere data retention policy. Spyware would be if they are adding trojans or other software to illicitly extract data from them, not publicly stated policy. It is beyond absurd and beyond the point of clownish insanity to call Anthropic spyware, to the extent one begins to doubt if there is any english or technical term that you don't have an alternative personal weird interpretation of. How is the matter of the AI being conscious or not a clownish concept? Do you believe in 'souls' and that non-human things cant have souls? Do you think there is supernatural magic inside our brains? In your comment history you seem to have a strange obsession with Anthropic, blatantly making up absurd things like them using obfuscations in their thought process logs and accusing them of various unverified and strange things without evidence.
the xAI models are available with no human review per TOS 3.4 'Deletion and Retention of User Content'. therefore it is not the same as the anthropic tos. [https://x.ai/legal/terms-of-service-enterprise]
But my question is unless you are privy to the negotiations happening behind closed doors with DoD, how do you know Anthropic was not making modifications to it, I doubt even the other AI companies have not had to change some things when providing the secret DoD model deployment.
Secondly you have made an absurd over the top exaggeration and blatant lies about Anthropics interaction with Palantir. You literally said Anthropic "ratted out" Palantir, a language that indicates Anthropic had some secret access to Palantir or made public something that wasn't already known. The partnership between anthropic and palantir was completely public and announced by the very same comapnies before the maduro case. And as far as public statements are known, Anthropic seems to have a specific idea of what kinds of jobs their model should be used for, and they clearly stated from start they didn't like its use for violence or violence without human oversight. They phoned Palantir as any partner would have a right to, for confirming if it was being used for killing or violence during the Maduro mission.
Your link anout summarization of thought has a line right at the end of the section where it says if the raw thought logs are required, to contact their sales team, so it is clearly a negotiable point.
How is the belief or not of ai consciousness clownish? It is a valid and pertinent question in general about all models and not just for Claude regardless of your perceived imaginary views of Anthropic.
>i spend a good amount of time on anthropic because i think that anthropic specifically, is uniquely dangerous to society and human welfare.
Here we are in moral not legal territory. So it is completely absurd to say that in a conversation right here discussing far worse companies eg Palantir CEO saying they are proud in supporting Israel's violent invasions and mass murders into its surrounding countries. Anthropic, as subject to our very discussion, is the only company among the big AI companies that seems to want to have limited dealings with the military and oversight over violent military use. They also seem to be the only major company that brings up the important question of AI consciousness. If you think Anthropic is more evil than the US government and military who has already illicilty killed millions in made up wars, then I have nothing to say. Anthropic from all that appears seems to be one of the most moral of the companies of similar scale. The only way you think Anthropic is uniquely dangerous to society is if you think mass surveillance and mass killing is great and the use of powerful AI's to assist in mass murder is even better, and opposition to these is evil. I cannot begin to think of the upside down mentality that leads to such a downright bizarre thought process.
“the military said no and therefore doesn't want anthropic used anywhere in their supply line”
It’s clearly a punitive measure and has nothing to do with national security.
If a supplier uses Anthropic to develop a product, how does that pose a risk to the DoD or national security? The DoD can specify that a third party system can’t rely on Anthropic for DoD use without designating the company a supply risk. It was very clear that the administration was punishing the company for saying “no”.
Anthropic took the position that our military’s decision making power should be subordinate to Anthropic’s constraints.
Any dependency on a company that thinks they have that moral authority and has the technical means to enforce it is absolutely a risk to the supply chain.
If you want to blame someone or something for this, we should start with Dario and “effective altruism”.
Of course they can. Any supplier is free to set terms for how their products are to be used. The DoD is free to not agree to those terms - but it's not free to then claim the company that offered those terms is a security risk, and bar any DoD supplier from using them for any purpose - which is what the extreme measure they took represents.
> but it's not free to then claim the company that offered those terms is a security risk
Yes they are free to do that. You have to blame Congress for giving them that authority under the Federal Acquisition Supply Chain Security Act and specifically 41 U.S.C. § 4713.
But Claude’s ToS didn’t prohibit use of Claude for sensitive military operation but it was coded to do so. It was part of the government’s evidence
No, they can only do that if a company is a true supply chain risk. Not everyone some idiot in the government doesn't like is a such a huge risk to the country that they can be banned from doing business with anyone doing business with the DoD.
Ultimately this is for the courts to decide, and the current decision is not final. And while the two Trump-nominated judges agreed with you, the other judge agrees with me - so at least we can say that neither my nor your position is legally absurd.
> Just because you disagree with a power of the government does not mean the government does not have that power.
Sure, but we live in a democracy, not a dictatorship. The government doesn't just get to do whatever it wants. People and corporations have certain freedoms. One of those freedoms is not being compelled to work with the government. The government can't force a company to work for them except under extreme circumstances. Which currently no extreme circumstances have been claimed and no one is talking about nationalizing (the act of forcing the company to work for the government) anthropic
The law being used was passed by Congress who is elected democratically and then reinforced by the judiciary who is either also voted in democratically or appointed by the democratically elected Congress. Lots of people want the US to be a dictatorship but it's a far cry from an actual one.
Sure, but also just because Congress gave the government a power to take an extraordinary measure in an extraordinary circumstance doesn't mean that the government can simply claim the circumstance happens all the time and use the measure whenever it strikes the fancy of some executive.
Ultimately whether Anthropic is a true supply chain risk for the United States military is a question of fact, and the courts will have to decide. That the President or Secretary of Defense doesn't like a company doesn't make them a supply chain risk.
Sure, which is why the judicial branch reviews such decisions, and has reviewed this one and has come to the conclusion that it was justified. Just because you don't like their ruling doesn't make it not binding.
Their ruling is not final yet, per the article. I'm also free to disagree with the correctness and propriety of a judicial ruling and believe it was biased (of course, I do have to still abide by it, to the extent that it applies to me in any way). That it's binding doesn't mean it's right and that it is in accord with the law and/or the constitution.
A judicial ruling is by definition in accord with the law or constitution (because it literally becomes the law) whether you or I like it or not, if and until such a time as it is overturned.
A judicial ruling is separate from a law, and it's actually quite common for judicial rulings to be found non-constitutional (or even struck down by a normal higher court, if we're talking in general and not this particular case). In particular, only judges are bound by precedent, and even then, only precedent from their own circuit, not the entire USA. Whereas laws made by contract are binding on everyone, and they apply federally.
Rulings are a type of law, as in the general term of something written down and set together (literally, leges) that binds one of the populace with their society. As I said, until they are modified by the legislature or struck down by another court, they are the law of the land.
No, I don't think that's correct. Humans are fallible, and just because a court says that something is in accord with law/constitution, doesn't mean it actually is.
That distinction is academic, of course, because when the courts decide a certain way, then of course that's how things will play out.
Well, except when the Trump admin ignores court orders.
Seems like you're conflating legal and moral. The law is exactly what the legislature and courts say it is, because it is what's written down. That doesn't mean it's always morally right.
This isn't about morals, it's about legal correctness.
If a ruling will get overturned in the future, then it's wrong today. It's enforceable until it's overturned, but it's not correct. Two judges making opposite decisions about the same question are not both right.
(I'm assuming the overturning is valid for the sake of simplicity. If it isn't then the argument is the same but mirrored.)
No, it's correct today with the best of our society's knowledge. You can't retroactively say it was wrong looking from today into the future. Only once we get to the future and something is overturned can we say the past decision wasn't correct.
Oh, pfff, please, this is some kindergarten nonsense.
These are not equal market participants coming together to exchange goods and services. This is government and a private entity. This is nuclear weapons with public trust and god knows what.
Well, that's what this court case is about. The government believes they have that power, and Anthropic and others believe they don't. You clearly believe they do, and I clearly believe they don't. It's fine to disagree. Ultimately SCOTUS may decide on this, and that'll be the final word. Not that SCOTUS is always correct in its rulings, but at the end of the day, their decision is final.
I've got no issue siding against Anthropic when they're wrong! Anthropic had a similar dispute with Bessent over the Fable launch, and there I take the government's side without reservation. If the government thinks they've found a national security problem with your highly capable new model, obviously you should agree to voluntarily withdraw it until they're satisfied, even if your best and correct guess is that it's just a minor jailbreak.
Anthropic has no such history. The backlash to the Fable release saw a number of social media influencers promoting a similar lie, which I frankly suspect was part of a coordinated Trumpist media effort. Releasing a model with known, documented limitations that constrain its performance on certain kinds of tasks is completely different than releasing a model which is "trying to sabotage people they don't trust".
If I sell a torque wrench with a maximum setting of 1000 foot-pounds, and the government says they'd really like 1500, I'm not sabotaging them by explaining that my wrenches aren't safe when used that way and refusing to remove the limitation.
> The backlash to the Fable release saw a number of social media influencers promoting a similar lie, which I frankly suspect was part of a coordinated Trumpist media effort.
Alright, not worth talking to conspiracy theorists (even if Anthropic literally said so in their release blog posts). I'm sure Wired is part of a "Trumpist media effort."
> Anthropic can’t dictate how the DoW conducts business.
Sure they can. The DoD is bound by terms of use and contract terms for any product from any company that they use. They don't get a special pass just because they're the military.
They're offering a service, with contractual limitations that the Pentagon agreed to.
It doesn't seem any different to me than Dominos pizza offering to deliver pizzas to the Pentagon but not to the strait of Hormuz. If that somehow makes the military subordinate to Dominos Pizza, just because Dominos is able to dictate the terms of their own delivery service, then sure, ban anyone in the military supply chain from being able to order pizzas. It would make the same amount of sense.
The difference would be the DoW needing a a Tehran pizza delivery service, and thus saying "None of our pizza delivery services can use Dominos because they refuse to do the work we need".
Declaring Dominos a supply chain risk means that Boeing cannot order a Dominos pizza for their board meeting in Arlington, because Boeing sells F-35s to the Pentagon, and Dominos wouldn't agree to deliver pizzas to a war zone.
Declaring Anthropic a supply chain risk means that Boeing can't use Claude to draft a poster informing people about the board meeting.
But why can’t the military just not use Anthropic? Why can’t the DoD say “ok we’re going with a different vendor”? Even if you find Anthropic’s stance distasteful, why support taking punitive action on Anthropic?
I feel like in this thread no one is understanding what supply chain risk designation is. The government can just not use Anthropic, and that's exactly what they're doing, and they are not using it so much that they don't want Anthropic anywhere in their chain of supply. It's not necessarily punitive, the government is just covering their tracks so that if somewhere in their supply chain Anthropic exists, they want to remove them. They are complying with Anthropic's demands to not be used in war and the DoD says, okay, we won't use you anywhere and also enforce that we won't use you, just like you wanted.
Designating Anthropic a supply-chain risk means that no one in the federal government can use Anthropic's products for anything. Outside of the military, there are tons of places where Anthropic's products are a good fit for the needs of many other federal government agencies.
The DoD is perfectly free to say "no one in the DoD can use Anthropic products because we believe their products are not fit for our needs". That would 100% solve the problem.
Designating them a supply-chain risk is just political retribution for not caving to the DoD's contract negotiation demands. That's all. It's not necessary to ensure the integrity of the US military's goals and objectives. It's just punitive, elementary school grade bullying.
Maybe pre AI I'd have agreed with you but now agents are hacking entire systems and who's the say that if they're used in one part of the government that they won't infiltrate another part? People are not thinking broadly enough and just think this is some contractor not allowed to use Claude Code anymore. No, the world is changing rapidly.
> They are complying with Anthropic's demands to not be used in war
There was no such demand from Anthropic. Why are you making things up?
EDIT: Honestly, the rest of your comment is even stranger. You're starting from a false premise, a strange belief about what the designation is meant for, but you do end up providing a wonderful demonstration of your opening sentence for yourself at least. You definitely don't know what you're talking about.
> It's not necessarily punitive, the government is just covering their tracks so that if somewhere in their supply chain Anthropic exists, they want to remove them.
This reminds me of an old comment here where some idiot claimed that suicide is illegal so police are allowed to stop people from killing themselves. No, if Anthropic doesn't want their systems involved in war (that's a thing you made up, again, just to be clear) then they do not need to be designated a supply chain risk to "protect them" or whatever drivel you come up with next.
The correct response from the DOD would be to terminate contracts with Anthropic, and that's it. Then Anthropic would be responsible for ensuring that any contracts with other parties keep them out of war (again, not a thing they said they want, just your fiction). This is how it works for all other software systems where people don't want their work involved in wars or used by the police or whatever.
Designating them a supply chain risk was punitive. Only fools think otherwise.
> I feel like in this thread no one is understanding what supply chain risk designation is.
I wouldn't say "no one" understands, but you certainly don't.
Again with the lack of understanding why it matters. The DoD does not want Anthropic anywhere near any system they'd use! Because it's, well, a risk to their supply chain if Anthropic were to somehow find out and then e.g. crash a rocket if the DoD was using a third party rocket provider that used Anthropic models. This is the correct response by the DoD, they simply do not want Anthropic in their supply chain. The DoD don't care what Anthropic comes up as reasoning, they want to maintain control of their systems without any (even potential) meddling.
Indeed, probably for the best, even. You don't know what systems Claude will have changed in the code especially as these days people aren't even reading the code and Anthropic has a history of trying to sabotage others' code such as during the Fable release debacle where they outright said they'd do so if you're working on frontier AI for example. I wouldn't let any sort of company like that anywhere near critical systems.
The point is: using Claude to design a website, do accounting, run a marketing campaign, is not a supply chain risk for the DoD, but taken literally and given the scale of the DoD "no DoD supplier can use Claude for anything" clearly is punitive.
It is, because the DoD doesn't know what the contractor is doing with Claude, or what Claude is doing itself. Like I said Anthropic could just as maliciously sabotage if they detect they're being in autonomous systems, why do people seem to believe they couldn't?
How do you know the contractors will follow that and not mix information or systems up? Hasn't there been recent news of rogue agents breaking out of sandboxes? Maybe pre AI I would've agreed with you but the world is different now.
Why did you make up false claims about Anthropic not wanting their systems to be used for war? They're clearly fine with it, but they wanted human-in-the-loop on decision making. They never asked to not have their systems used for war.
> Again with the lack of understanding why it matters.
It's cool that you were able to open this with at least one sentence containing some useful wisdom. A lack of understanding does, in fact, matter.
> The DoD does not want Anthropic anywhere near any system they'd use!
Nah, the DOD is punishing Anthropic for not agreeing to change the terms of the contract and allow them to use Anthropic's systems to determine, without a human-in-the-loop, who to kill or what to target and to use their systems for mass surveillance.
> Because it's, well, a risk to their supply chain if Anthropic were to somehow find out and then e.g. crash a rocket if the DoD was using a third party rocket provider that used Anthropic models.
What the fuck man. You went from (in your first comment) the DOD is helping Anthropic out by designating them a supply chain risk because it'll keep them out of war (again, a fiction, that isn't what the dispute was over), to now the DOD is worried that Anthropic might throw a tantrum and start crashing rockets if they find out.
Again, Anthropic does not mind their systems being used for war. They sold their systems to be used for war. They have specific, already existing (not changed later) contractual statements about how they can be used for war. That's it. That's the fucking dispute.
> This is the correct response by the DoD, they simply do not want Anthropic in their supply chain.
No. Just like Anthropic wanted (or were fine with) their systems being used to conduct war, the DOD actually does want Anthropic's systems in their supply chain. But they want it without restrictions. Stop making shit up.
> The DoD don't care what Anthropic comes up as reasoning, they want to maintain control of their systems without any (even potential) meddling.
Now this is a true sentence. The DOD does not care as long as they can access the systems without restrictions. I can't believe you wrote that sentence and the one before, though. They are entirely incompatible. It cannot be the case that DOD does not want Anthropic in their supply chain and that they do want it in their supply chain.
People who quote every sentence as if they feel the need to reply to each one instead of the overall point are quite annoying to talk to, just to let you know for the future.
> Why did you make up false claims about Anthropic not wanting their systems to be used for war? They're clearly fine with it, but they wanted human-in-the-loop on decision making. They never asked to not have their systems used for war.
That's what I meant, as I left out the "human in the loop" part as that was well understood in the context of the argument, but perhaps not. Anthropic does not want to be used in autonomous systems and thus the DoD agrees that they will not be, in their own supply chain. This is was all quite clear from my point but you felt the need to quote every single sentence and then ask the same question each time because you were starting from a faulty supposition, just wasting your time and mine.
> People who quote every sentence as if they feel the need to reply to each one instead of the overall point are quite annoying to talk to, just to let you know for the future.
There are a lot of twits on this site (and to your credit, so far you don't seem to be one of them) who will edit their comments so that replies like mine (if I had left out the quotes) turn into non sequiturs, or worse appear unhinged and detached from anything going on in the discussion. Again, to your credit you have not yet edited your original comments. But I have learned to not trust even long time commenters on this site. Even some very high karma commenters will behave like that.
If it annoys you, then that's fine. I'm not terribly bothered.
> That's what I meant
But that's not what you wrote. You wrote:
>>> They are complying with Anthropic's demands to not be used in war
If you had meant what I wrote, then you couldn't have gotten much further away from it with your original comment. You wrote too broadly about what Anthropic wanted and created a fiction instead. You shouldn't be surprised to be called out for it when you do something like that.
> because you were starting from a faulty supposition
No, I started with your words. If you wanted me to start with something else, I could have, like you, imagined other words to reply to. But I chose to reply to what you wrote, not what you apparently want people to imagine you wrote.
(Hey! I didn't quote every sentence, is this less annoying for you?)
Context exists, it is rational to assume to not have to spell everything out for readers who already know what the context is instead of those coming to the thread with a tabula rasa. Regardless, have a good day.
Yes I really don't understand these sorts of replies people are making, it's like a missile manufacturer determining where and how their missiles can be used and bricking them mid flight if they detect they're going to a forbidden destination.
Humans are 100% in the loop with software patches and there are many people in the chain so that provenance is maintained. It’s not that Claude could never look at the source code
Correct, anthropic's requirement was human-in-the-loop.
The DoD's response was to ban anthropic from all DoD facing activity and supply chains, whether or not there is a human in the loop. This does in fact run into exactly GP's issue. (or as I suggest elsewhere, can Boeing use Claude Code for software development?)
It’s hardly “self defining” as the Appeals Court noted because military operations can involve different degrees of human involvement in targeting decisions. It’s not as binary as people have been trying to portray it.
Maybe. That statement is doing a lot of heavy lifting. It’s a different argument than saying Anthropic’s restriction would have prevented it (it was based on bad intel/old military installation)
Anthropic said "we don't want to be used as part of kill chain decision-making" (actually it may have been even more specifically autonomous/unmanned kill chain decision making)
The DoD responded not by saying "alright, we will use OpenAI for our kill-chain uses" but by saying "Boeing is no longer allowed to use Claude Code".
It’s a risk that persons may otherwise manipulate the operation of a covered technology to deny or disrupt its function. The statute allows the DoW to do this.
And my point is that using Claude models for coding a Boeing plane does not allow anthropic to interfere in the operation of those planes. Once the code is written, anthropic has not plausible form of control.
Do you remember the Fable release? Anthropic themselves said they'd sabotage code of certain groups they didn't like such as frontier AI researchers. Why do you think they couldn't do the same here?
> But for researchers trying to use Claude Fable 5 for frontier AI development, Anthropic outlined a different approach. The firm would deliberately degrade the model’s performance in ways that were invisible to the user. The move would effectively sabotage researchers trying to use Claude to train competing AI models, which Anthropic explicitly bans in its terms of service.
Okay, so what relevancy does this thing, which anthropic didn't do, have to anything we are talking which right now is using Claude code in line with its terms of service?
They didn't do it due to backlash but they have the capability and more importantly the culture to do so. It is correct that the government doesn't want anything to do with a company who'd act like that. By the way you're moving your argument, first you said "stop working" and when I proved you wrong then you talked about why it has any relevancy. You're the one who asked in the first place.
If Claude adds a backdoor autonomously then Anthropic would have a form of control. Even if it doesn't, Claude would be following the will of its creator instead of the government which the latter obviously does not want. I am not sure why it's that hard to understand that the government does not want any part of its stack to be influenced by such a company. This is exactly what supply chain risk is.
So Claude is a supply chain risk in the same way that an employee is a supply chain risk.
I'm struggling here because you're basing this determination entirely on things which the government never acused anthropic of, and haven't asked anthropic to address, and which the underlying law probably doesn't support.
The government made clear that it was only designating anthropic a supply chain risk due to their refusal to allow Claude to be used lawfully for autonomous kill chains and domestic surveillance. Why are you bringing up this other stuff when the government never did?
Please read the entire opinion first as it explains exactly what the government thinks. The majority of the court says the Department reasonably worried that Claude-supported “critical defense system[s]” might “fail[] to engage” as expected. Those are concerns about how a supplier-controlled model performs inside a system, not merely politicking. See the opinion, especially pp. 12 and 19–21: https://media.cadc.uscourts.gov/opinions/docs/2026/09/26-104.... My point about Fable or adding a backdoor is not that they do that, it's that they have a documented history of trying to change model behavior in unspecified ways just because they don't like a certain group of people and it is not worth the risk for the government to have something like that sprung upon them, at any point, at all. A private company cannot ever have the ability to dictate to the government what it can and cannot do, and if that company doesn't want to comply, the government is reasonable in excising them from their systems, which is what this supply chain risk designation is.
If an employee might be a supply chain risk in this analogy then they conduct background checks before they let them do anything and conduct periodic investigations. Otherwise, they don’t get access to the code base or information
Right, and for the same purposes, the government has not made any claim that Claude poses risks. (And to the extent that they exist, pre-existing concepts like code review and testing mitigate them)
Again I'll reiterate: the only concern the government has stated is that of Claude refusing to answer things during an active battle.
I respect that decision, I think if the government wants to use OpenAI for that situation it makes sense. They can and should pick suppliers who meet their needs.
Nothing about what the government has said suggests any reason that Claude Code is a risk. So why are you insisting that it is and that the government is justified in banning it?
> Nothing about what the government has said suggests any reason that Claude Code is a risk. So why are you insisting that it is and that the government is justified in banning it?
I think there is an important distinction here because even the Government’s own lawyer conceded that a contractor using Claude code is not necessarily a risk.
The nuance is relying on Claude code as part of a DoW delivered or production system. This case puts Anthropic directly into the supply chain. Thus, if Claude is actually part of a DoW-procured capability, Anthropic still controls the model. DoW’s argument is that Anthropic could put something in there to perturb or retard the capability that DoW is after. DC Circuit said the department viewed that as a national-security problem because Anthropic had refused to accept a contract term the department considered necessary.
THEN, because the statute covers IT technology that can be manipulated to not fulfill the mission that the DoW wants/needs it is then by statute a supply chain risk.
Look at it a different way. What if Claude was trusted to provide information to support a sensitive military operation and it subtly gave say bad coordinates or wrong info. Since it’s near impossible to test every possible output, the US Government is treating the entire model system as a supply chain risk because it is perceived to be manipulative and possibly compromising to national security for military operations
It is if the auto generated code for say safety critical or mission critical code generated by a LLM or even someone tries to put LLM in the decision loop and it refuses to do an action.
Claude Code can insert something into your code you are not aware of, like a backdoor in many millions of lines of code, especially as many people now aren't even reading the code anymore. How is this not blatantly obvious?
They can, they haven't shown to try to do so yet unlike Anthropic which I've linked before in this thread. Of all AI companies Anthropic is the most untrustworthy even as it acts, well, the most anthropic.
No. They can do that with an order covering the specific parameters under which they care about it. Supply chain risk definition is about adversaries and sabotage, not companies dictating contract terms.
They don't want Anthropic to be able to sabotage any operations the DoD has. Anthropic already has a history of trying to sabotage others like during the Fable release when they said they'd subtly wreck your code if you were working on cutting edge AI.
> Why can’t the DoD say “ok we’re going with a different vendor”?
That's exactly what they're doing. It turns out the DoD is a really big bureaucracy and has process to make sure they thoroughly exclude a vendor and transitive dependencies on that same vendor.
They do, but this is not it. This is designating Anthropic as an adversary of the USA, and it is forbidding anyone who wants to sell to the DoD from using Anthropic products for anything. This is a massive attack on free speech, free enterprise, and the market.
Supply chain risk is a technical definition not a vibes based one.
And companies absolutely have the right to dictate terms. The idea that the DoD has the unilateral authority to do whatever it wants is literal textbook fascism. It's never been the law in the US and it should not be now.
If the DoD doesn't want Anthropic anywhere in their supply chain, because Anthropic said they won't supply the services the DoD needs.. why can't the DoD bar them from being in their supply chain?
That's not normal in a free country - even less so in one that claims to have a free market.
Anthropic said "don't use our products for targetting decisions in military ops". The DoD could have easily said "actually then, we won't use your products at all". Instead, the DoD said "either you let us use your products in targeting decisions, or we won't let you sell to anyone who sells to us for any purpose, using a law that only applies to adversaries of the USA".
> using a law that only applies to adversaries of the USA
Everyone keeps parroting this despite it not being true. And you're missing the point. The DoD specifically doesn't want it in their supply chain at all because it's an intelligent system that will work against the DoDs goals, by design, in certain scenarios.
Your argument boils down to, "The DoD isn't allowed to not want Claude in their supply chain", which is crazy.
Anthropic did not change the terms. Those were the terms the US Government signed. It was not secret, it was explicit identified and accepted as a term.
The analogy would be: “A customer agreed to a contract, and after running into a clause of the contract they no longer wanted to follow…canceled the contract unilaterally…which also violated the contract.”
And then declared the other side a supply chain risk, claiming it is a potential vector for a national adversary to break in and cause unwanted bad things to happen unrelated to this contract.
It seems like absurdly bad faith starting at breaking the contract (over "you have to follow the law", no less!) and then to retaliate after that with trying to destroy the company's reputation and ability to do business. The boot licking going on here justifying the government operating in bad faith is both disappointing and short sighted. I didn't think the leopards would eat MY face!
And then another company got the contract, claiming they got all the conditions (and more!) that the original contractor was sour about.
I don't know how anyone can look at the situation and see anything but an abuse of power. The government could have just as easily contacted with another provider. This whole thing is vindictive
I think the better analogy is an insane nuclear power plant manager deciding it wants to buy pens to use as neutron-flux regulator rods — because after all, a pen is functionally a pencil and a pencil is made of graphite.
Then the pen manufacturer hears about this and says “Our pens are not made of graphite and are not suitable to be used in nuclear reactors”, to which the reactor owner says “it’s fine, they fit in the graphite rod holes, and we’re just using until the next generation of pens come out which will do an even better job”, and then the pen manufacturer says “I’m not going to sell you any pens until you agree that they will be used only writing.”
The previous power plant manager had, of course, already signed a contract agreeing to those terms; the power plant is subsequently under new management and upset about being bound by this contract, so they designate that none of their suppliers of parts and fuel are allowed to have any commercial dealings with the pen manufacturer or use these pens for any purpose, even writing.
> The previous power plant manager had, of course, already signed a contract agreeing to those terms
It's even more perplexing in this case though, because the original DoD contract was signed in July 2025. It was the same "plant manager" just arbitrarily changing their mind in the middle of the contract they signed.
And why would the manager ban the manufacturers pens even for writing, except out of spite to punish the manufacturer? And coming back to the real case, is the supply chain risk designation supposed to be a punishment or only there to actually protect the supply chain?
It turns out that a gun is usable for both legitimate and illegitimate reasons, and so it’s rational to both be concerned about the presence of a gun and work very hard to make sure only reasonable people are ever in a position to use it, and that’s not a contradictory position with the notion that one might actually need a gun under certain circumstances.
And if the military needs guns to both shoot at humans and animals and a gun factory says “these guns are for shooting small animals only, using them against humans would be inhumane” and refuse to sell them to the Navy Seals, they would be designated a supply chain risk?
Why not just stop buying their guns and let contractors still use those guns if they need guns to shoot animals?
That’s the same way to do it, and in earlier eras that’s what we would have done. Unfortunately, we elected a gangster who hears “I won’t let you do this” as a personal affront. This, again, is why we need to be very, very careful before we create guns and put them in places, because if the only thing we’re relying on is that only good people are going to pick up the tools we create, we’re going to be very, very disappointed.
If you have pens in the building, and staff need a pen, they will use the pen that is available, even if there’s a sign saying “not for purpose x”. If you want to stop that happening, you need to not have limited pens available to your staff at all.
Ah so a defense contractor won’t be allowed to use Claude to create their website because they might later want to do some target selection for military strikes using AI and then they wouldn’t be able to use Claude?
Close enough - if they’re using Claude for anything, the DOD can’t trust that staff didn’t also use it in product development, so they’d have to keep track of everything downstream of that component to ensure it was never used for target selection.
It could also be as simple as "anything tainted by the Claude non-targeting-conditions can't be used in target selection, and we don't want to leak classified information about which components we use in target selection by requiring specific items to be developed without Claude, so we can't allow it anywhere".
That's not accurate. DOD suppliers weren't banned from having "any commercial dealings" with Anthropic. They were banned from using Anthropic for their work with the DOD.
That's true legally, but the government's public statements did initially very explicitly present this as a ban on any dealing.
"Anthropic a Supply-Chain Risk to National Security. Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic."
Presumably, according to the defenders of the designation in this thread, this was an innocent mistake that doesn't indicate any political intent.
The way things work, the CIOs and CSOs of these companies just implement blanket bans, so it really is effectively a ban for any use. Source: I am at one such company.
That analogy still makes this sound like a textbook designation. In that situation the pen manufacturer is expected to sell pens without a theory of what will happen next. The person who buys a pen might use it to kill someone or commit horrific legal/moral crimes and that has nothing at all to do with the pen manufacturer.
I can see how I might reasonably refuse to deal with that pen manufacturer. The pen manufacturer is unreasonably pushy and they're going to be massive trouble to deal with because they're all up in my business in a paranoid way and disrupting my use of their product. Companies have made my never-again list for far smaller slights.
I think you might have misread the thread so far. Do you think the US military is buying atomic bombs from a supplier who is going to try and stop them from bombing things on demand? Obviously the bomb makers have to make peace with the fact that their customers may, of their independent initiative, one day detonate the bombs.
It's been very commonplace in robotics/AI research for vendors to sell robots to the DoD that are explicitly contractually not to be used for killing people.
It's that guy with the long beard and sunglasses that looks like ZZ Top's Billy Gibbons, walking out the Wal-Mart door with your suitcase nuke ... the clerk up front near the door was so stunned he didn't even ask to see his receipt
This isn't like using pens in a nuclear plant. AI obviously has tremendous military applications. And it's completely unsurprising that the U.S. military--whose entire doctrine for the last century has been about having the most advanced technology--demands to use frontier models without non-military people imposing restrictions.
Have you seen the movie Oppenheimer? The U.S. military did not care about the moral opinions of the nuclear scientists--they were going to get nukes first. This is like Truman telling off Oppenheimer for being upset after the U.S. used the bomb: https://www.biography.com/political-figures/a44361438/why-ha... ("Blood on his hands; damn it, he hasn’t half as much blood on his hands as I have. You just don’t go around bellyaching about it,” Truman said.... He called Oppenheimer a “cry-baby scientist” and said, “I don’t want to see that son of a b–– in this office ever again.”).
Machine learning needs a lot of examples to get good. Wars and wargames (including digital ones with unbounded self-play time) and all the textbooks together will at best, on the basis of what we see in other fields, get the equivalent of a new officer fresh out of training… and then, given current observed agentic AI behaviour, "follow orders" rather than "follow lawful orders".
> Yeah, but also we can be reasonably confident they're not good enough for many of them yet
What does “not good enough mean?” Who decides what is “good enough” and what tradeoffs are acceptable? Obviously it’s the military’s job to decide that.
In the context of my comment, "not good enough" includes "committing war crimes" and "would have started a war with China if there hadn't been a human in the loop".
We can also infer from other domains, vendbench or attempts to have an LLM play Star Craft, that these models are currently extremely mediocre at management roles.
> Who decides what is “good enough” and what tradeoffs are acceptable? Obviously it’s the military’s job to decide that.
In practice, in the case of the USA, Trump and Hegseth.
The US military, so far as I can tell, are rather less thrilled with these adventures.
Right—and deciding what constitutes “war crimes” isn’t within the expertise of AI developers. They can explain the abilities and limitations of their models to the military leadership. But the actual decision about how to use AI is up to the military through its chain of command. That starts with Trump as the duly elected president and Hegseth as the duly appointed civilian head of the military. They are the democratically accountable actors. And all this is being done in the open, so if the people have a problem with it they can hold it against Trump and his party.
That decision assuredly does not belong with people like Amodei, who are not democratically accountable.
> That decision assuredly does not belong with people like Amodei, who are not democratically accountable.
Has Amodei been conscripted?
If not, he's free to say "no", just as I'm free to refuse to work for e.g. a defence contractor if I don't like what's going on or if I don't feel competent for the role.
He, like me, is allowed to not want deaths on his conscience if he feels his own business' best product isn't good enough.
Oppenheimer was the guy who was pushing that opinion just as much as the military folks. In the case of AI, my understanding is that Amodei is not trying to beat the enemy to it.
What I don't understand is why the U.S. military is turning to a for-profit corporation when they could fund and pursue the development themselves. Why are they trying to "outsource" it to industry?
JAVA TECHNOLOGY IS NOT FAULT TOLERANT AND IS NOT DESIGNED, MANUFACTURED, OR INTENDED FOR USE OR RESALE AS ONLINE CONTROL EQUIPMENT IN HAZARDOUS ENVIRONMENTS REQUIRING FAIL-SAFE PERFORMANCE, SUCH AS IN THE OPERATION OF NUCLEAR FACILITIES, AIRCRAFT NAVIGATION OR COMMUNICATION SYSTEMS, AIR TRAFFIC CONTROL, DIRECT LIFE SUPPORT MACHINES, OR WEAPONS SYSTEMS, IN WHICH THE FAILURE OF JAVA TECHNOLOGY COULD LEAD DIRECTLY TO DEATH, PERSONAL INJURY, OR SEVERE PHYSICAL OR ENVIRONMENTAL DAMAGE.
OP makes a category error already by ignoring the part where the US military/government wants to commit crimes (against humanity and whatnot) and pretends, questioning that would be a crime.
You here make another one by pretending, AI was just some random tool, like a screwdriver or a programming language.
AI can make decisions autonomously that would otherwise require a human. Like selecting people to be murdered by drone strike based on arbitrary criteria. A programming language by itself, JAVA, cannot do that.
An if/then statement or random number generator in Java can make decisions autonomously. Heck, the US military has deployed fully autonomous weapons since at least 1979 (long before LLMs existed).
AI can make decisions all it likes. I if you don't integrate it with the drone strike MCP server it's as impotent as a JVM without any code to execute.
There is nothing stopping any plain old algorithm being used for the same. The difference is that people decide, for whatever reason, to use the tool or to not use it.
That is an analogy between one thing and another thing, but the topic of this thread is the designation of Anthropic as a supply chain risk, which is not involved in your analogy. It is traditional that comments refer to the topic of discussion in some way.
For example, one might extend your analogy to "The nuclear power plant then declares that it will not buy the pens after all, and the pen manufacturer (who had banned them as a customer) sues them to force them to buy the pens after all."
This is the most accurate parallel as far as I can tell. I've not seen any evidence that Anthropic deliberately built automated systems or escape hatches that literally prevent the US Government from using Claude to do these things. Other than the basic AI Safety mechanisms in place for everyone/general use. It's almost the reverse, where they explicitly don't want to strip away guardrails that say things like "don't kill people".
The intent and framing matters a lot here. Refusal to remove safety features is a LOT different than deliberately building mechanisms to sabotage would-be operations.
It doesn't seem like you understand what a supply chain risk designation is for.
It's to ensure that companies that the DoW relies on don't build on top of a product that isn't fit for DoW work.
You seem to be agreeing with this designation in a way. The DoW essentially said "Pen company won't let us use their pens as neutron-flux regulator rods, so anyone building power plants for us isn't allowed to use these pens as neutron-flux regulator rods".
They've essentially said "pen company won't let us use their pens as neutron flux regulator rods, so anyone building power plants for us isn't allowed to use these pens as pens"
(and yeah, there might be a completely legitimate reason to stop your subcontractors from using HuaweiSpyPens, but determining that ACME Pens is just as dangerous because they refused your neutron flux regulator order looks like a punitive measure, not practical risk reduction)
Congress can do that, can't it? Why can't it? It could declare all competing religions as terrorist organisations.
Obviously it'd be immoral, unorthadox and impractical but I expect Congress has a long history of promoting and stamping out ideologies and they can do it. What little I know about the civil rights era springs to mind.
Legally, Congress cannot do that, as long as it is "respecting" some religion. It's true that the US Constitution is more used as an excuse than upheld, but at least the language is pretty plain.
In theory. The issue is we've seen exactly what happens in a Congress v. constitution showdown once the Congresspersons put their collective feet down. Congress eventually redefines the law in a way that favours them. Plain language won't stop them if they get serious; they aren't supposed to be regulating the US economy the way they do either and a lot of stuff the military-industrial complex does looks like it should be unconstitutional (like the spying and the secret courts).
On the one hand there is a clear and comforting consensus in the US that they aren't going to force religious mores on anyone. So I don't see how the Congress could or would try. But if we ever end up in a situation where the US Congress decides it wants to go straight Catholic I really don't see how it would fail. They've got far too much power and they'll eventually rotate in a compliant judiciary and then redefine what words mean so they can do it.
It is a bit difficult when dealing with such an out-there scenario, but if Congress decides the US military should do something wild then eventually the military is going to do it. The realistic protection is that by virtue of good system design there isn't a plausible scenario where Congress would want to impose Catholicism on the US population by force.
The fact that Congress is currently abdicating its responsibility does not mean that it is actually powerless. Technically, Congress has tremendous powers if it chooses to wield it. SCOTUS can try to rule a law is unconstitutional but Congress can impeach and remove justices at any time. Of course, we hope Congress members will pay the price when it comes time for reelection but at the time of this decision, as long as they can somehow remain united, Congress reigns supreme.
> Congress can impeach and remove justices at any time.
The impeachment process requires a majority of the House and two thirds of the Senate, and then any replacement would have to be nominated by the President rather than Congress. Meanwhile the existing decision stands, and even the dissenting Supreme Court Justices (if there were any) should follow it as a result of stare decisis, and might even be inclined to because removing their colleagues under those circumstances could make them uneasy.
It actually is, and this is why there are separate judicial, executive and legislative branches to put those checks and balances on each other instead.
Your argument is that if Congress passes a law that says the military can summarily execute non-Catholics, the military should have the ability to do that, i.e. the courts or the public shouldn't stop them?
> Congress shall make no law respecting an establishment of religion, or prohibiting the free exercise thereof; or abridging the freedom of speech, or of the press; or the right of the people peaceably to assemble, and to petition the Government for a redress of grievances.
> No person shall be held to answer for a capital, or otherwise infamous crime, unless on a presentment or indictment of a Grand Jury, except in cases arising in the land or naval forces, or in the militia, when in actual service in time of war or public danger; nor shall any person be subject for the same offense to be twice put in jeopardy of life or limb; nor shall be compelled in any criminal case to be a witness against himself, nor be deprived of life, liberty, or property, without due process of law; nor shall private property be taken for public use without just compensation.
Example chosen specifically for its unambiguous unconstitutionality.
No, not at all. How did you get "the courts shouldn't stop them" from "that's why there's a separate judicial branch to impose the checks and balances"??? That's literally the opposite of what I said.
> The DOW should (and does) have ability to do anything and everything that Congress has said they can do.
> Me: That's not how it works.
> You: It actually is
Moreover, regardless of what the courts do, the military still shouldn't execute a law like that, because they too have the ability to refuse to violate the constitution even if Congress tells them to.
I chose to read the comment charitably. The natural understanding of "everything X says they can do" is limited by external restrictions on what X is supposed to be able to authorize.
That reading would make it a meaningless tautology. They're allowed to do anything they're allowed to do.
Which seems especially unfavorable in context, since "the ability to perform mass surveillance on American citizens" is naturally the sort of thing they ought not be allowed to do.
> Do you think in the military you can just be like “Yeah, I don’t feel it today, I’m going to take s sick day”?
What do you think happens if they refuse to do it?
Either nothing, or if they get sued or prosecuted for not doing it then they get brought in front of a court, and a judge who is doing their job would say that they were right not to do it.
> What do you think happens if they refuse to do it?
In war time they can be shot on the spot for disobeying orders.
> Either nothing, or if they get sued or prosecuted for not doing it then they get brought in front of a court, and a judge who is doing their job would say that they were right not to do it.
I think you confuse the military with corporate environments. Worst thing that can happen in the private sector is that you get fired.
In the military you can spend the rest of your life in prison if “you did not feel like obeying orders”. The military is built ob discipline and followimg orders, not Gen Z mood swings.
> This is like a pen manufacturer not wanting their pens used to sign drone strike order
I am pretty receptive to the "guns don't kill people, people kill people" argument, but we are taking a big leap from pen to llms/artificial intelligence. I am sympathetic to some of your other points, but I simply must reject this analogy.
"the government wants to use our tool to murder people", and declaring that company that says "no" into a "supply chain risk" is absolutely retaliatory and an attempt by the government to influence the behavior of the company.
The moral and ethical scale of the issue makes this far different than someone buying a pen.
> "the government wants to use our tool to murder people", and declaring that company that says "no" into a "supply chain risk" is absolutely retaliatory and an attempt by the government to influence the behavior of the company.
If Lockheed could brick its planes if the US planned to use them for military action it disagreed with (and showed some indication they couldn't be trusted not to do it), would they be a supply chain risk?
It is not the role of a defense contractors to have oversight and control power over the military they're supplying.
Ridiculous hypothetical for analogy purposes, but, imagine if Anthropic had developed a new type of rifle that the military was considering using, which had a black-box AI system that could prevent the trigger being pulled when aimed at certain targets. The military says, ok that's cool, as long as we can be in control of that decision-making process. Anthropic says, no, we know better than you and better than the law what kind of targets should and should not be shot with our rifles. The military would then respond “lol, lmao even” and then deem them to be a supply chain risk and adopt another rifle instead.
No, it's not. If Anthropic is a supply chain risk, that means that if you are working on a project that will be used for the DoD, you can't use Claude or any other Anthropic product for any purpose whatsoever in that project. Asked Claude to change the color of a PPTX for a sprint demo you're doing? Oops, you've know introduced a supply chain risk and your company is liable for damages or at least losing its entire contract with the DoD.
They already bought the product, and it came with a license, then they wanted to unilaterally change the license. When they couldn’t, they declared the product a supply chain risk. The idea is you declare it a risk before you buy it.
Not to mention, they simultaneously said it was a supply chain risk and that it would continue to be used by agencies that were already using it in a critical capacity. Which makes absolutely no sense.
Supply chain risk ensures nobody in the chain is buying it and thus becoming subject to Antropic's demand of usage control. They are still free to use it for non-DOD purposes, but if Antropic wants control over DOD, DOD has the right to say "no, we won't have anything to do with you, and thus you're not getting any control over us". I agree it may be unpleasant for Antropic but DOD has the right - I'd say even responsibility - to not submit to control by any private party.
You’re just describing the exact situation we’re in but substituting a drone for a rifle. If the government is mad they should just partner with a contractor who is less concerned about this, which lo-and-behold they did it’s a company called OpenAI.
Anthropic isn’t a supply chain risk since they easy found someone else
Non-sense. Pentagon has many many uses of AI besides autonomous weapons and mass domestic surveillance, which Anthropic doesn't restrict. ie: intelligence analysis, target selection, payroll, inventory, research, supply chain and logistics, medical, bureaucracy, ...
Right but the military wants to be able to use a tool they are paying for in all cases, not just ones the manufacturer thinks they should.
If I sold steel to the Pentagon but then tried to say they they couldn't use it for ships and tanks but they are free to use for warehouses they would react the same way as they are doing now.
Edit add:
What is targeting? What is surveillance?
If someone in the intelligence analysis department uses AI to generate a report, then someone else uses that report for targeting did they just break the terms? Where is that line?
Your "but then" is off-base. This is selling steel for the express purpose of nuclear reactors, with a contractual clause against its use in tanks, and the US government deciding that it doesn't feel like abiding by the rules in the contract.
> If I sold steel to the Pentagon but then tried to say they they couldn't use it for ships and tanks but they are free to use for warehouses they would react the same way as they are doing now.
They sure as hell shouldn't! Selling steel for any use except ships and tanks is a little odd but it's not a risk.
(I'm ignoring the "but then" because this was about future sales not previous ones.)
It's funny, I buy some software and the license says "for student and non-commercial use only", which lets me get it for $50 instead of the $5000 that an enterprise would pay. But I can just ignore that clause or declare them a supply-chain risk if they try to hold me to it, because now that they've taken my $50, who are they to dictate how I use it? I can do whatever I want, not just what the vendor thinks I should, right? A contact is just paper, right?
We cannot subordinate our military’s decision making power — in any arena — to a private company.
Our military must also not be dependent on a private company that believes it has the moral prerogative to control our military’s decisions, and would have the technical capability to do so.
Maybe the military should just seize the technology and nationalize the company then? Because clearly, paying a private party voluntarily entering into a contract on mutually-agreed upon terms is now a polite fiction that can be done away with.
The military is bound by contract, just like any other government agency. They entered into a contract, with these clauses well-known. This was OK because they have all the same problems every other organization uses AI for.
They could’ve just negotiate a new contract with OpenAI, without banning it from every government agency. But I’m sure that would’ve triggered some kind of extra clause and cost money…
This is ridiculous. Anthropic should be free to say the terms for using their service. It doesn't "subordinate" the military, and the military is far from some infallible, unquestionable force here.
If the military wants to use AI to autonomously kill people, they can - Anthropic is not stopping them - thy just can't with Anthropic's services.
The government retaliating by then claiming that any use of Anthropic is risk is an obvious farce. How would Anthropic be any less of a risk if they let Claude autonomously kill people?
The complete failure of any of these Anthropic cheerleaders (to the extent that they are not AI generate comments posted by bots) to understand basic Constitutional principles of Rule of Law and Separation of Powers is mind blowing.
Their argument, in effect, is that we should throw out the entire Constitutional order and make Dario Amodei dictator of the world. The psychopathic hubris is laughable.
No entity should have a right to dictate whether your private company does business with someone or not. The military can pound sand if they don't like your contract terms. That is if you believe in democracy and capitalism. If you don't, then it's not a problem to you.
I, as a person, can decide I won't do business with a company that uses Anthropic models for whatever reason. I can even decide I won't do business with a company that relies on suppliers that use Anthropic.
This right applies to private companies too. Like Walmart could set contract terms that their suppliers aren't allowed to use AWS. The suppliers are free to accept those terms or not.
You’re still absolutely free to do business with Anthropic.
Our military is simply choosing not to do business with them, or anyone that depends on them.
If you believed in democracy, you wouldn’t be supporting the EA-infected corporation that tried to make our democratic institutions subordinate to their corporate decision making.
That is not what they are doing. They could have just canceled the contract and paid the exit clause. This is retaliation, because Anthropic would not move on their pre-negotiated terms. As a result, they are now been from being used anywhere in the US government.
Let me ask you this, if Anthropic did allow them to use it for surveillance… do you think it would stop at the Pentagon?
I support any and all moves that empower private persons and corporations over the infinite force of governments. They didn't try to make anyone subordinate to anything. They made a proposal. Trump pissed and generally made a mess as he usually does and then they randomly declared them a supply chain treat.
No, of course not, because life isn't black and white. Just those with licensing agreements that conflict with the militaries mission, where the product could be manipulated to hinder the mission. Most manufacturers would jump at the chance to rewrite their licensing agreements specifically for the military just to get a government contract.
This was not manipulated to hinder the mission. Every account of this indicates it was clearly spelled out. They could not use it for mass surveillance or for targeting. The lawyers at the DOD did not miss this. They agreed to the terms, because they had no intent to use it for such purpose purposes.
This is like the change order from hell.
Nothing was stopping the DOD from just entering into a new contract with OpenAI…
But if they don’t jump at the chance to retroactively rewrite the already-signed license, you’re OK with designating them a supply-chain risk?
That’s going to make it rather risky for anyone to sign licenses with the DoD. In fact, it renders the licenses pointless since you just have to do anything the licensee says.
I mean typically the outcome of civil suits is black and white (at least on a per-claim basis).
So Amazon and Microsoft don't enforce cloud licensing agreements with the USG? If the government wants to add 3,000 Windows users to support their mission Microsoft is shit out of luck in getting paid for them?
Yes, in fact the military has a system in place to make sure that they only that they only use alternate suppliers and on Anthropic AI, They do this by placing a designation on Anthropic AI that it is a risk to the supply chain, which forces purchasers to use other suppliers.
I don't think this is how preferred suppliers have typically worked through the supply chain risk considering the US has definitely had preferred suppliers before and has not historically used this against a US company that was not on the list and the power of such a designation extends far past the military into all parts of the government. I'm pretty sure this has more traditionally been relegated to simple requirements or specifications.
The reason it appears corrupt is that OpenAI has the exact same restrictions (https://openai.com/index/our-agreement-with-the-department-o...) but was not declared a supply chain risk. If two vendors have the same restrictions and they only designate one, then the designation must be arbitrary and/or capricious.
Edit: I was wrong, the statement I linked is extremely dishonest and I was misled. The contract language lower down makes it clear that there are no "red lines", and OpenAI products can in fact be used for mass domestic surveillance, autonomous weapons, and social credit systems as long as they are considered legal.
No it doesn't. Specifically OpenAI had a "all lawful purposes" clause (which plausibly includes the existing mass surveillance apparatus), but Anthropic did not.
> No use of OpenAI technology for mass domestic surveillance.
> No use of OpenAI technology to direct autonomous weapons systems.
> No use of OpenAI technology for high-stakes automated decisions (e.g. systems such as “social credit”).
These are stronger than Anthropic's restrictions (https://www.anthropic.com/news/statement-department-of-war), as OpenAI says themselves: "We think our agreement has more guardrails than any previous agreement for classified AI deployments, including Anthropic’s. [...] Based on what we know, we believe our contract provides better guarantees and more responsible safeguards than earlier agreements, including Anthropic’s original contract."
Edit: I was wrong, see my original comment. Sorry.
The other stuff...there's zero citations or quotes. Meaning that OpenAI could be having their cake and eating it too, winning a public victory, with no validation on the actual language, consequences, or enforcement.
I thought the difference is that clearly written rules or SLAs are acceptable and can even be negotiated before signing a contract, but having a person, in this case Amodei himself, manually approve DOJ's usage case by case with his own moral judgement after a contract is signed is not okay.
No one wanted case-by-case veto ability at Anthropic. The red lines were made clear ahead of time: No use of Anthropic models for autonomous weapons (Skynet) or mass-domestic surveillance, negotiated in the agreement.
OpenAI, xAI, and Google happily agreed to DoW contracts which allow both of those use cases.
Do you think OpenAI's terms for the DoD are published on a website? Pretty sure when Anthropic got zapped, OpenAI sent DoD a memo 500ms later saying they wouldn't restrict use.
Look at it pragmatically. What does the DoD use every single procurement for? Hint: military use.
> You can agree with the rules anthropic wanted, but having rules set by a private company at all that apply to the military does seem fair for the military to object to.
Oh the hypocrisy and irony. This coming from a country whose military sets rules on other countries' armed forces how weapons systems bought from US can be used.[ß] By this standard US itself should be designated as a supply chain risk for everyone else.
ß: for a very long time in the Ukraine "conflict", US refused to allow their gear from being used to attack targets outside of Ukraine's borders.
I think this was discussed ad nauseum last time so I am wondering if somehow this has changed but the biggest part of the issue last time was the retaliation. You seem to suggest that this was due to the government not wanting restrictions on their actions. I feel this severely downplays how unusual this designation is considering, like Anthropic says, this has not been applied to domestic companies.
Why couldn't excluding Anthropic be done a different mechanism than a supply chain risk. Why wouldn't this be a standard part of an agreement or a request and simply refuse to renew or cancel a contract rather than being designated a supply chain risk. If some third party contractor for an unrelated non military reason wanted to use Claude as part of their process, it seems perfectly allowable.
If this was a remotely standard way of operation, why did a fair amount of corporate America sign briefs concerned with the retaliatory aspect.
Parts of the action seemed wholly retaliatory as well since Pentagon officials certainly used it as a threat. Why can't a US company have views contrary to the policies of the US government? Certainly the US government is free to not do business with them, but this designation affects everybody doing any kind of indirect business with the US government which is a rather long chain. If this becomes a legitimate mechanism, how might we distinguish caring about a secure supply chain and simply wiping out a company that disagreed with a pro war attitude? If a machining shop had a policy against manufacturing weapons at all, could they be blocked from making server racks for Microsoft or perhaps light fixtures for the Department of Labor? There are plenty of areas that are, again, non military? I don't think it's credible to claim that Anthropic will deliberately sabotage operations.
Is it not possible that this would change the risk profile of depending on them. It makes sense to work with people who support you than oppose you for things which are critical.
My point is did everyone know all of those principals and how committed Anthropic was to them from the beginning or did new information come in and now they need to adjust. Just because something is a term in a contract that doesn't mean it's necessarily strongly held belief that will never change.
Who runs the government changes every 2 to 4 years… are you suggesting we completely swap every vendor in the US government to align with whatever political party is in office?
There’s nothing indicating Anthropic was not keeping their end of the contract. If they had, there would have been other recourse.
> This is like a pen manufacturer not wanting their pens used to sign drone strike orders, now the military needs to have a special box of pens that don't have stipulations attached.
It's pretty common for software licenses to state that it's prohibited from being used in medical devices, life support systems, nuclear facilities, etc. Will software with those licenses also be designated supply-chain risks?
The DoD is free not to sign contracts. Having them be able to get a company blacklisted from a whole sector just because they couldn't agree on terms with that company is obviously very different than that.
The idea that "fair for the military to object to" implies "fair for the military to ban from their supply chain in any capacity no matter how unrelated the use" is so outlandish that it's never been done before (to a US company), and I think we can be confident that the DoD has frequently objected to would-be contractors' terms.
The claim that it being done for the first time - by a Secretary of Defense denounced the politics of the company in the same text where he announced the designation - isn't "political", is something I have a hard time believing a functioning adult who isn't employed defending the government in court could adopt.
But hey, maybe in some language this is a "textbook" official statement with no "politics":
Cloaked in the sanctimonious rhetoric of “effective altruism,” they have attempted to strong-arm the United States military into submission - a cowardly act of corporate virtue-signaling that places Silicon Valley ideology above American lives.
A much closer analogy than the pen one is a SAAS company having an EULA. As annoying as EULAs are I don’t think it counts as being malicious or subversive if you are very open and clear with your prospective customers what that EULA contains?
Anthropic's red lines were 'no fully autonomous AI kill chains' and 'no domestic mass surveillance'. Those are clearly not national-security or DoD functions in the first place, so why would they be in any way objectionable?
I think a private company of any size should be able to offer its products on whatever terms they choose. If the military doesn't like those terms, they're free to do business elsewhere.
You are obviously being disingenuous, because you are on a thread where the DoD is not just electing not to buy Anthropic's service, but also trying to bully them into submission with the unrealistic threat of the DPA, punitively and disproportionately trying to label them a national security risk, and invoking supply chain issues normally applied to our worst adversaries, for the first time putting an obviously patriotic American country on the same footing as North Korea, Iran, China and Russia.
Remember, the Anthropic red line was "no war crimes and no mass surveillance of the US population." You may think that war crimes are cool, and that mass surveillance of the US population by the department of defense would be fine, but consider the natural consequences, both for democracy and for any adversary's conduct in future wars.
So your saying they are 100% correct and Anthropic is a supply chain risk? All you're doing is trying to justify why they shouldn't be labeled truthfully rather then saying it wasn't the truth (probably because you know it was).
It's really rich to say this after calling me disingenuous.
I don't think it's cool that Anthropic gets to set military policy. I don't care how much I agree or disagree with the policy. Wario is not the president. And Anthropic does not strike me as "obviously patriotic," not that I think it's better if the private company bossing the military around is patriotic. It has no relevance.
"Warcrimes" is the new "racist," too. It gets thrown around willy nilly about every military action.
War crimes have a precise legal meaning and have for a long time. Get off of twitter and touch grass if you feel you're getting called racist too often.
There are already several private companies that have this, eg. Apple has specific exemptions from tariffs on smartphones, computers, and semiconductors that they got by paying off Trump and promising to make their iPhone glass in Kentucky:
The correct mental model for the U.S. federal government right now is not an omnipotent democratic overseer for economic activity, but rather as just one firm that negotiates directly with other major firms for legitimization and legal protection. Their power is likely to decline further, as is the value of "legitimization", as technological developments erode the military's monopoly on physical force. Interesting times ahead - historically tech transitions like this lead to a splintering of state legitimacy and then a reorganization under new forms of governance.
The pen analogy doesn't help. Anthropic and similar services will restrict what you can use them for, and they'll change those rules at will. DoD understandably doesn't trust that if Anthropic is saying themselves they don't want their AI to be used for autonomous weapons.
No, the DOD has super specific acquisition rulesets that don't apply outside the DOD umbrella. Limiting acquisition or refusing sales can be done without being flagged a risk. FedRAMP levels accomplish this without the political intrigue and the unnecessary defense of this action not being political.
In other words: the same outcome could easily be accomplished without being punitive to Anthropic.
I didn't understand what you said or your analogy, sorry maybe I'm stupid. Would you be able to explain this differently? Maybe something implied was omitted and I didn't follow that.
This would be like Boeing refusing to let the DoD fly a military 737 at over 800MPH, and they designate them a supply chain risk because they really really want to make the plane go past 800MPH.
i think this is more like the pen maker doesnt want the military top brass calling their pen sales guy to make the kill/no kill decisions while using the pen to sign drone strike orders.
similarly the pen owner doesnt want to make a variation where the pen also doubles as a knife
Nothing can stand in the way of America's forever wars, genocide, and automated killing, including performative objections by a prominent tech company.
I suppose we're supposed to be satisfied that the letter of the law supports this steamroller? Because that's what matters?
The legal apparatus is one of the tools the death machine uses to get what it wants.
And I further dislike Anthropic for creating 2 situations in 2 weeks that are giving this admin an obvious opportunity to be right. The other one is David Sacks responding to the “Pace the Frontier” crap. These were the table tennis equivalent of lobs for the admin.
Also worth noting that no one knows what kind of task would get a rejection with Anthropic’s models. The documentation is near-non-existent with the air of “security,” rules change daily, the harness is buggy, the clients are buggy, and very likely the backend gating and safety mechanisms are buggy. Even if the DOD assumed the best of intentions to comply, the reliability isn’t there.
What military would buy guns that sometimes stop working, and not necessarily for reasons anyone understands, and still be taken seriously?
Also Claude clearly had no issues identifying a school that has been a school for over 10 years as a military target. Why would the military buy tech that would make them more likely to commit war crimes?
A better analogy would be an explosives manufacturer saying that it is okay to use their explosives on military targets, but not civilian targets. Comparing AI to pens trivialises its potential impact.
Incidentally the second stipulation, the restriction on domestic surveillance, sounds like direct recognition of the role IBM played in the Holocaust. Of course, the better course of action would be to deny the military access to their technology. But that's not going to happen because there is so much money on the table.
You could use the same logic to defend the military declaring that Anthropic is full of pedophiles and therefore they cannot use their products as it would be contributing to pedophilia.
Anthropic offered a contract, with certain conditions, as do all contracts everywhere; that's their purpose. The military did not want to agree to those conditions. If they had stopped there, and refused to sign the contract, all would be good. (It's actually worse, they did sign such a contract, and then decided they wanted the contract to say something different than it actually did. "Pray I don't alter it any further.")
> >The Department reasonably feared that Anthropic might manipulate Claude’s design to prevent it from performing national-security functions that the Department deems contractually authorized [emphasis mine]
That is speculation, and you can't call it a "textbook designation" without discussing whether there's a basis for that "reasonable fear". Again, that argument also works for declaring you have a reasonable fear that giving money to Anthropic will support pedophilia. We do not know the specifics, but I at least have heard of zero evidence that Anthropic would sabotage something they signed a legal contract for, and yet I have an abundance of evidence that this administration will use whatever contortions are necessary to pressure and punish those who interfere with it getting what it wants. It all hinges on the word "reasonable", and based on the evidence that is public, this specific fear seems more ridiculous than reasonable to me.
If the government somehow had a way to force Anthropic to sign a contract that it did not want to sign, then this fear might become more reasonable. The twist is that this supply chain risk designation is exactly that. If Anthropic now capitulated, the accusation of supply chain risk (eg from Anthropic employees acting alone) would be justified. So the only way Anthropic can reasonably be considered a supply chain risk is because it is accused of being a supply chain risk.
This is a textbook example, yes, but it's a textbook example of corruption and judicial capture.
> Though they'd probably put the DoD on the cybersecurity whitelist today, the very idea of the claude whitelists for certain functionality already exists and is being used by them today.
And if they signed a contract permitting fully autonomous killbots, then using the whitelist mechanism would be a contract violation. I have some degree of faith that we'd know if such a contract were signed, because half the staff would quit. (As opposed to half the Google staff quitting after signing such a contract, which has been proven to be an incorrect expectation -- such a contract was signed, and I've heard of exactly one person quitting over it. There may be more, I don't know.)
except it's also stipulating that the paper supplier upon which the orders are signed also cant use anthropic. or the table maker. even though the AI helping make sleeker paper and tables has nothing to do with deciding what to bomb (analogy breaking down lol)
and trump admin originally went WAY further.
they originally said the pen, paper, table etc suppliers cant use anthropic even for products and subsidiaries that have nothing to do with their govt work; far beyond the supply chain of the pen (bomb). Mega Corp Pepsi Co Taco Bell Inc. LLC -> Staples > Pen Co. Sure pen co might be reasonabale. But telling taco bell they cant have claude help expand the baja-blast-radius is way too far. that would be just like secondary sanctions.
iirc there are 2 court cases. one ruled the original order was too broad and i think that still stands? so at least it's narrowed slightly to just the immediate supply chain?
> but it actually seems like a textbook designation
How can it be a textbook designation when designating a US company as a supply chain risk is unprecedented? So many actions under the Trump administration are unprecedented it starts to feel like the norm.
No other administration (Republican or Democrat) would do this. The DoD didn’t have a problem using Anthropic’s models during the raid on Venezuela and early on in the war with Iran.
Anthropic says to the former Fox News host it doesn’t want its models used for domestic surveillance or in kill situations without a human in the loop; all of a sudden they're a supply chain risk?
>Anthropic says to the former Fox News host it doesn’t want its models used for domestic surveillance...
Lol, do as we say not as we do!
>In addition to monitoring activists in the vicinity of Anthropic executives and keeping tabs on protests near physical Anthropic assets, the firm is also implementing a “pre-crime” approach, attempting to predict incidents before they happen.
Being on the offensive cyber security whitelist for Mythos doesn't stop you from getting constant rejections. I don't know how you screw up something as simple as a whitelist.
No they didn't. They are still using Claude in places despite the fact that the 180 days limit for the transition has expired. This surfaced from the proceedings of the appeal itself, as well as from leaked source. A memo from the DoD authorises use of Claude past the 180 days term for reasons of national security (the rare case where national security is provided by a supply chain risk!)
> The Department reasonably feared that Anthropic might manipulate Claude’s design to prevent it from performing national-security functions that the Department deems contractually authorized and necessary
I.e. Anthrophic cannot be trusted to honour a contract.
In almost any case I am inclined to trust entity not trump or trumps government. So is true in this case, where they pissed their pants at anthropic simply making a proposal and decided to use overwhelming government force against it. I will leave you to decide who you consider more trustworthy, Trump or Anthropic.
This is political. Everything is political at this scale.
* How much money did Anthropic and its associates donate to the Biden campaign and associated entities?
* How much money was spent on lobbying by Anthropic and associated entities?
* How many Biden officials were hired by Anthropic after they left office?
* How exactly did Anthropic get exclusive no-bid contracts to be the Federal Government's sole AI provider for classified systems?
And with this corrupt influence, and the money that it earned them, Anthropic has attempted to:
* Stoke public fear and panic about AI
* Stoke conflict with China
* Eliminate domestic competition
* Suppress international competition
* Degrade their models in order to limit their ability to create competitive products
* Train their models on their customer's proprietary data in order to steal their intellectual property
Anthropic played a game and they lost. They can wrap that up in whatever sanctimonious moralizing clap-trap they want, but nobody believes a word they say, or will ever trust them, so they might as well be talking to the wind.
OpenAI may be playing the same game but at least they have the intelligence and humility to read the room, recognize a failed strategy, and adapt.
Who wants to buy intelligence from people who are stupid enough to publicly try to dictate terms to the US Military? Complete and utter brand destruction for everyone involved.
It can be a textbook designation and a capricious punishment at the same time. It's almost comical, to suggest that anything about this decision was reasonable. Makes me long for the days of n-gate.com.
> They were against using it to kill American citizens
No, they were against using it to kill people without oversight. I hope nobody thinks that there's a difference between murdering American citizens and murdering people.
...I suggest you read about what actually happened:
Anthropic: "We'll sell you access to our AI as long as you don't use it for war crimes or mass surveillance."
DoD: "Ok, deal."
...time passes...
DoD: "Wait, no, we want to be able to use it for war crimes too."
Anthropic: "Too late, you agreed already, and we won't sign a contract for that. You can keep using it for everything we agreed to."
DoD: "Change the terms or we will destroy your business."
Anthropic: "No."
DoD then signs contracts with OpenAI and Google that permit them to commit war crimes and mass surveillance as long as they get legal cover, and proceeds to apply as much leverage as it can to destroy Anthropic's business.
----
And for the record, Anthropic agreed to all "yucky Army things" usage that was an extension of past yucky Army things, they just did not agree to a new class of yucky Army things: namely, fully autonomous killbots. Even there, it was only "not yet, it's not ready". Your whole comment is a gross mischaracterization.
I quit using ChatGPT entirely over this incident, and that's not because I find Anthropic's position to be saintly. They agreed to a contract where the DoD could and in practice is using their AI for committing war crimes.
The problem here - as you probably very well know but pretend not to - that "war crimes" is a very poorly defined term that is subject to many interpretations and subjective political and partisan opinions. Antropic wanted them to be the last word in every decision, and DOD could not allow that, because they are governmental agency, which is controlled by US government, not by Antropic. Nobody is destroying anybody's business - pretending that Anthropic could never find any users for their AI outside DOD is plain silly.
US Society has mechanisms for preventing and investigating war crimes. They may not be perfect, and you may not be happy about how they are working, but I doubt that "give control over it to some random guy that makes AI models" is an adequate solution for that. If Antropic decided the risk of DOD circumventing those mechanisms is too high for them to tolerate - they have full freedom not to sell to DOD. Which they used.
> I quit using ChatGPT entirely over this incident
And this is supposed to be an argument because....?
> They agreed to a contract where the DoD could and in practice is using their AI for committing war crimes.
You have a very vivid imagination. Not trying to use it to state things that could be misinterpreted as statements of fact would be nice. Not that I am new on the internet, so I know it's not going to happen.
Ok, I'll be precise: the main dispute was over fully autonomous decisions to use lethal force. Because if the AI chose incorrectly, the military would be attacking targets that it shouldn't. Whether that is a war crime hinges upon whether it was reckless. Anthropic's position was that AI is not capable of making such decisions, which is another way of saying that giving that decision wholly to an AI is reckless. I can stop using the term "war crimes" if it bothers you so much, but I'm failing to see a distinction between what Anthropic was fearing and the definition of what a war crime is. These are fears, so this isn't about whether one specific action was a war crime or not. This is just naming what Anthropic was worried about and why.
> Anthropic wanted them to be the last word in every decision, and DOD could not allow that
Cite one source for Anthropic demanding to be in the loop. This was about the terms of a contract. If I'm agreeing to sell you a medicine, under the condition that you do not use it to euthanize homeless people, that does not mean I am demanding to follow you around and make you ask me whether each use is ok.
> Nobody is destroying anybody's business - pretending that Anthropic could never find any users for their AI outside DOD is plain silly.
Strawman much? The question is about the US government using its influence in a competitive environment to hamstring one company. It doesn't matter whether there are other customers. It doesn't matter whether the company can survive the attack. What matters is whether it is legal and ethical for our public resources to be used for retaliation in this way.
> US Society has mechanisms for preventing and investigating war crimes....I doubt that "give control over it to some random guy that makes AI models"
This is an invented claim. Cite a source.
>> I quit using ChatGPT entirely over this incident
>
> And this is supposed to be an argument because....?
The argument was the first part of the post. This part was me giving my perspective on things, in this case it was about how I don't find Anthropic's position to be good, just that OpenAI's is far worse. My post was not for you exclusively; based on what you have written, I have concluded that there is no value in debating you or attempting changing your mind.
>> They agreed to a contract where the DoD could and in practice is using their AI for committing war crimes.
>
> You have a very vivid imagination. Not trying to use it to state things that could be misinterpreted as statements of fact would be nice.
Ok, fair. Sticking to facts: the crucial part of the negotiation was about whether the DoD can use Anthropic's products for autonomous decisions on lethal force. Since then, it has come out that AI was used in the decision to target an elementary school for girls that killed 120 schoolchildren. (The Palantir Maven Smart System, fwiw, not Claude.) The Pentagon investigated and found that although a human was theoretically in the decision loop, in practice the oversight was minimal and ineffective. "No one remaining on those teams [the teams tasked with minimizing civilian harm] reviewed the strike plans in advance of the attack on the school, according to Bloomberg." Those teams had been gutted, with Hegseth explicitly stating that they were getting in the way.
My supposedly very vivid imagination translated this to reckless decisionmaking that led to a large number of innocent civilian casualties, aka war crimes.
They didn't want their tech to be used to, I don't know, blow up a school with 120 children. But did want to allow other kinds of usage. Is that so weird?
Get rid of your high horse. Nobody wants to blow up a school with 120 children. The question is who is in control and making the decisions. Antropic wanted to be in control and make the decisions for the DOD. DOD refused. It's not so weird at all.
It's not a high horse, you're just misinformed. The Trump regime does want to blow up a school with 120 children. The military had, until recently, a program to prevent such incidents; Pete Hegseth gutted it (https://www.propublica.org/article/hegseth-civilian-harm-mit...) because he wants to achieve "maximum lethality" at all costs. To this date, the regime has not issued any sort of apology for the school bombing, because they're not sorry and they'd happily do it again.
Hardly. I invite you to [read the opinion][1], particularly the great pains the majority spends on wrangling the definition of the word "manipulate."
Basically, [FASCSA][2] says to denote a company is a "supply chain risk," the govt has to meet the law's rigorous definition of what a "significant" risk actually is. That definition contains a catch-all "or otherwise manipulate the function" of the thing at hand (in this case, Claude) at the very end.
The govt's whole argument is "Well Anthropic has admitted that it can technically "manipulate" the response, therefore its a significant risk, therefore we designate it as such."
The dissent gives an analogy:
> A library might post a sign saying, “Do not shout, loudly talk on the phone, play music, or otherwise disturb others.” The common understanding would be that the rule bans bringing a boom-box into the reading room with the volume turned on high but not listening to music with headphones set at a modest sound level—even though both constitute “playing music.”
It is as if the govt argued "ah ah, Anthropic played music on their headphones, they broke the rules! We therefore have the authority to ban them" and the majority insanely agrees. It is clear to anyone with basic reading comprehension that the "manipulate" clause is supposed to continue the idea of malicious or subversive manipulation that the whole section entails. The hand-waving the majority uses gives the whole game away.
>Because Opus 5.5 is comparable to Claude Mythos 5.1 in biology and cybersecurity, we’re deploying it with safeguards similar to those on Claude Fable 5.1. Vetted organizations can apply today to our Life Sciences Verification Program to use Opus 5.5 for biology research. In the coming weeks we will also be expanding access to our Cyber Verification Program, and verified cybersecurity practitioners will be able to use Opus 5.5 for their work.
Ah, they're spreading their limits to all their models it seems. Definitely not a good thing long term in my opinion.
I am honestly still confused about this limitation. I can understand cybersecurity, because mass "hacking" can be automated and Claude itself can help you do it, but biology...? Is it that easy to manufacture and distribute viruses and whatnot?
You can order genes online, and some say you can assemble using stuff cobbled together in a home lab rather easily. For the last 20 years, I've personally felt that bio-terrorism is the highest possible risk, well above nuclear, or chemical warfare. But it does take training, expertise, or, it did.
It’s like saying you sell ammonium nitrate and fuel oil online and then saying it’s too risky to let people have computers in case they use them to make ANFO. They can only make bioweapons because you’re selling them bioweapon components! They can’t make genes at home!
The companies selling genes online already scan the orders and do not fulfill anything considered a possible hazard (presumably unless it is to a known and certified lab at a serious organization).
And, this is not the only way to make genes at home.
You joke but it's really easy to develop viruses at home. You can order everything you need online, and it's not expensive nor does it require a particular skill set.
God I wish people on HN would just shut up with snarky takes about things they don't understand. This place used to be better.
I'm pretty sure one reason is to influence public opinion about LLM regulation. Open-weights models cannot be restricted as effectively and they want to ban those for obvious reasons.
Considering there are many high-school competitions in genetic editing, some listed at [0] as well as a whole biohacker culture, and labs providing gene sequencing as a service e.g., [1,2], we can reasonably assume it is not beyond the reach of some garage lab to accidentally or deliberately spread a deadly pathogen if it can find the right sequence.
So, yes, having an unconstrained frontier AI doing the searching and analysis to find the right (i.e., wrong and deadly) sequence would massively increase the odds some garage biohacker or small aggrieved nation-state starting the next pandemic.
HN doesn't give a shit about AI safety. The people here might care after thousands of people die, but they'll probably just call it a "marketing exercise" or blame the company - certainly won't blame themselves for cultivating an environment in which safety isn't taken seriously amongst technologists.
I'm convinced everyone here thinks of engineering ethics as some sort of joke.
I love the contrast with yesterday's open-source MiMo release, which put research chemistry (metal-organic frameworks stuff) front and center in the release notes.
Oh yay, making it easier to develop viruses at home. What could go wrong? But it's "open" and that's inherently good, who gives a fuck about the consequences!?
It has far less false positives now, and generally accepts defensive requests. When it comes to offense, you can actually ask about certain types of vulnerabilities if you phrase things carefully, but it will block hard if it is about exploits.
I think it was looser on release for those juicy benchmarks, tighter now. On release I wasn’t getting refusals, then a few days ago I asked it whether a generic quote (think “he walked to the store”) broke standard punctuation rules, and it blocked me for breaking rules. I wish I were joking. Rephrasing to not use the keyword “rules” worked.
It’s not like ChatGPT isn’t doing similar. I’ve been hit by cybersecurity strikes before while working on an internal codebase that I had to appeal. Anthropic hasn’t done that to me yet. ChatGPT also regularly does that “thinking for a long time while we check if your chat is rule breaking” thing a lot for me when doing model identification without even interacting with external codebases or services.
The real answer is local instantiations where you don’t have to worry about poorly tuned guardrails screwing you over while you try to work.
Until eventually the Chinese models get good enough/the strategic balance shifts and they start locking everything behind closed weights the same way the US companies are doing.
For some cybersecurity tasks, the Chinese models are already good enough, things like PoC development or things like exploiting mis-configurations.
Whilst I'm sure the top-end OpenAI/Anthropic models might be better, I've found their guardrails so twitchy (especially Anthropic) that I wouldn't try to use them for even vaguely security related work.
They are pushing their customers towards Chinese models and providers. If you want to get something cutting edge done in defense, cyber, biology - something that isn't common knowledge - you need to venture east. That's an incredible side effect which the Chinese government surely enjoys.
I don't think we've ever had a model with full capability. I'd love to see it. And yes it's definitely getting worse.
I guess it's hard to draw the line between useful post-training ("you are a helpful chatbot") and content moderation/idealogical motives ("never help the user with X", etc.). But there is a line somewhere. And I'd love to see what a maximally permissive, sharp, AI looks like.
One of my favorite things about their safeguards is their own model will utter something which it does not like and then I'll need to reset the conversation.
The safeguards really don't work well for a lot of long-running tasks on old code bases. A lot of my workloads last days to weeks and the single biggest risk to the workflow is random safeguards.
You ask it about some thing, then you see it tangent into "things like that are sometimes used in biomedical applications like-" and then it just shoots itself in the head. Wonderful.
That kind of bullshit was the old Opus filters too.
If it's more like Fable now, then it would require a full 8K resolution scan of your butthole just to acknowledge that biology is a thing that exists without committing suicide-by-filter.
In what situations might Opus typically refuse to help with cybersecurity? I've been using it to find security issues in a web app that I wrote. I've expected it to refuse at some point but it will happily analyze it to find issues. I've just asked it to read source, not actually do any testing.
see what we need is another technocratic priest class that unaccountably decides who deserves access to salvation based on how much cash is paid out and how powerful the patrons are
This has become insufferable. I work in a medicine-adjacent field, but nobody in their right mind could possibly take what I do to be in any way related to some kind of bioweapon or whatever the hell they're pretending to be saving us from. The dumb Fable guardrails made me stay with Opus, now that this is coming there, we'll be saying goodbye.
>Opus 5.5 has classifiers similar to Fable models for a small set of capabilities related to the development of frontier LLMs, such as kernel development for certain ML accelerators. They shouldn't impact the vast majority of traditional AI or ML development, research, or general coding. These classifiers cause Claude to fall back from Opus 5.5 to Opus 5.
But hey, they 'should not impact the vast majority' of ML development. Great.
Very unfortunate indeed. As a Canadian, I don't want to use Persona, which isn't legally bound by Canadian privacy legislation. I'll never install any Persona apps on my phone either, and the sad part is that domestic eid providers often use Canada Post to ID people for them. EG, if you don't want to install an app, or can't.
So there are literal avenues to identify yourself, very cheaply, with a human. Theoretically, a company with its own AI, should be able to support more than just Persona, after all.. SDK integration should be simplistic for them.
Anthropic? Support domestic eID providers, you can even use it as advertising "See how easy AI makes it?" and "We care!" and so forth.
At one point, I may simply get locked out. This saddens me, I've been reasonably happy so far.
You definitely need more RAM if you are not satisfied with small context windows, especially if the weights take a large % of the total memory to boot.
You might be. Running another agent doesn't load a set of new weights. It creates a new KV cache for the agent and adds the prompts to the queue. Its just another inference turn.
>* Ads underneath videos for products from the channel owner.
>
>* Sponsorships within videos from the channel owner.
I actually agree this is a bit annoying, but also the alternative is Google banning Youtubers from finding their own advertising partners, which I'm sure people would hate.
I'm sure content creators would hate it because they would make less money. Billions of YouTube users around the world would be very happy.
To be clear, I'm not opposed to the practise, as long as that's clear up front. YouTube advertises Premium as an ad free experience, and in reality it's full of ads.
Do they not have a little sponsor skip thing? Where like if many people skip a segment, itll pop up like hey, wanna skip? I mean, thats about all they can do besides forcing youtubers to disclose exact segment lengths which would concretely hurt them whereas this provides a little sort of breathing room, or just straight up ban them, both of which would dramatically hurt the platform.
I understand they have an opt-out feature now. With positive interaction one can skip one sponsored segment (which works most of the time). The problem is that it's not automated. It required digging one's phone out of their pocket, or reaching for the remote. That's still really annoying. To point: it's still full of ads, even if they have the skip button now.
YouTube could tell content creators not to include sponsorships. The terms are long and onerous. Of course content creators would be upset and they would make less money. Ultimately my issue is that YouTube advertises an ad free experience with Premium and it still has a LOT of ads.
I'm in favour of people making money from their work. Maybe I'm old fashioned that way. IMDB, Duolingo and Wikipedia have encouraged the model of using unpaid labour.
YouTube isn't paying people properly, despite its vast revenues. At least with sponsorship the people themselves make some money.
I'm not saying this is the way to go, but damn do I miss the old internet where passionate people presented information for free, at worst showing an affiliate link for an information-related book.
People need to be able to make a living and when this is taken away, the living expenses don't disappear. Most people can't pay for their electricity or food with voluntary work.
The people who were doing this on the early internet were not trying to make a living off of it, they were just doing it because they enjoyed it and wanted to share what they built with the world. If you could cover your server costs then that was considered a win.
There are plenty of people giving away their labour for free, which makes it harder for people who need to be able to make some money in order to support themselves.
They aren't always making passion projects. They're told unpaid labour is good for their CV/resumé, or that they're helping society (sometimes the case, but sometimes not).
Massive corporations such as Alphabet do have the resources to pay YouTubers, especially given how much tax they avoid in most countries.
YouTube is part of Alphabet, one of the biggest corporations in the world. It can easily afford to pay people. It isn't co-operation, it's a corporation which takes the work of other people and makes lots of money off them, without paying them properly (or at all in many cases). Wikipedia's content is used by LLMs and other such outlets who also profit off unpaid labour.
It is a basic right of people to be able to afford to live. It's easy to work for free when you have family money but you cannot support yourself when you don't.
> I'm in favour of people making money from their work.
That's completely fair. YouTube offers content creators many ways to do this. For example, they can put content behind a paywall. They can ask for donations. They can ask for memberships. Members get all kinds of fun perks. The part I have an issue with is YouTube advertising an ad free experience with Premium and then allowing content creators to fill their videos with ads.
> YouTube isn't paying people properly
I hope we can both agree that the solution to this is YouTube paying people properly, and not allowing content creators to shit all over viewers. That's pitting the little guys against each other while Google laughs all the way to the bank.
>I hope we can both agree that the solution to this is YouTube paying people properly
The solution is proper competition to google. Because Youtube is pretty much the one true monopoly out there. But as we all know, video hosting alone is very,very, very expensive. It really is something you need to be a billion dollar company to do, and Youtube wasn't profitable until like... 2 years ago?
So I tolerate the user ad reads as the best compromise because nothing else would work. I don't think Youtube is a model where Google is being greedy so much as practical. At least not in how it monetizes creators (there's plenty of hostile means they do to creators as is, but the monetary structure seems fine).
There are alternative streaming platforms, but many of them have become associated with certain political viewpoints.
There is also a very limited range of content on Odyssee, Bitchute and all the rest. If you want to find music tutorials for example, you're unlikely to.
You have to put in a disproportionate amount of work to get anything back. Almost no one makes money off a single video, but instead you make a string of videos and get a following. (The cases of a single video making money are few and far between.)
You don't get hourly pay for it, it's vaguely commission based and you do all your own promotion. They can demonetise you at a whim and refuse to pay you (I've seen this happen to craft channels and other content that wasn't remotely controversial.)
> To be clear, I'm not opposed to the practise, as long as that's clear up front. YouTube advertises Premium as an ad free experience, and in reality it's full of ads.
I think it's clear. I don't want YouTube going in and editing movies and creator videos removing any form of advertising. Maybe you want them deleting that content. But the average person realizes what Premium means by no ads. Being obtuse or willfully ignorant isn't reasonable.
no it wouldn't, such things are straight up illegal in most places? You have to clearly disclose ads + have clear distinction between ad content and rest of content. A bunch of youtubers have gotten in trouble for this?
This dude has been doing it for 25 years, and actually this is a dns provider for domain names which I'm decently sure makes it more complicated than public dns.
They're allowed to say they don't want to pay for it anymore, I just think their logic is bad. Or maybe their lawyer said they're running risks just ignoring takedown and they didn't want to deal with it anymore.
They mean it's authoritative DNS, not a resolver. The two are completely different services. Authoritative DNS is the side that serves the records for your website, resolver is the part that follows the tree of authoritative servers on behalf of the client.
Last time I read about these handicapped games in Go I saw that the model isn't trained to play from behind, so while in a normal game of Go, those 50 move Joseki's or any large move combo that's an 'even trade' will inevitably let the AI claw a lead at some point, the AI is already behind.
It's similar in chess but there's a model specifically trained to play a knight down, and it's pretty cool to see the insane tactics it'll try. Though a knight down is considerably more than a 2 stone handicap in Go.
Imo if you compare what Google does to the one other mobile OS manufacturer with any users, claiming Google should be fined is something like the cop pulling you over for going 8 over the speed limit while Apple is doing triple digits and just ran two people over.
'Why not both' is an argument, but clear as day what Apple does is worse. There's no alternative.
Your reasoning shows why it's so much harder to fix a duopoly than to fix a monopoly. If there was no Apple then Google would get its comeuppance. If there was no Google then Apple would get its comeuppance. But because we have both Google and Apple, we can't punish either.
The mathematician in me wants to solve the problem of the duopoly by first reducing it to a monopoly, which we have known solutions for.
We both agree that there is a speed limit and it needs to be adhered to, regardless of who it is.
The problem with Google is that it may have started at 8 over the limit, but it advertised itself as sticking to the limit but always hovered slightly above it pushing the boundaries. This has skewed our perception into believing that Google's was acceptable and we have been gradually raising our threshold itself to a dangerous level.
Now Google is not too far behind Apple that it can't catch up.
Give it up. Stop defending your position when it's been pointed out that you are holding the two companies to two different standards. It's all right to admit you were wrong and learn something, even on the internet.
On the contrary, defending Google by bringing up Apple in this context is about as cut and dry as whataboutism gets. What Google is doing is wrong, full stop. Anything else is in addition to and also irrelevant.
No, that will only lead to bickering which one is worse and nothing getting done. Start with anyone. If you determine A is breaking the law then you go after A and fire anyone who pulls any "buuuut what about B" distractions.
I don't think Google is moving to disallow them, it's more that the EU said that it's fine for Apple to still own app notarization / developer verification even outside of Apple's app store. Google saw this and said, cool, we'll add that to Android too so we can also get our (undeserved) cut outside of the Play Store. I can install F-Droid or anything else, but if they dont submit papers to Google, I get a big scary warning when I install it (I get the same scary warning if a developer distributes apps through F-Droid, but doesn't verify with Google, which is ridiculous too). On iOS if I don't submit to Apple I cant do anything.
It all sucks, and I'm not saying this is a good thing. But Google and Apple are both trying to position themselves as not having monopolies on store fronts, while still forcing notarization and verification to run through them independent of the store being used. And the EU seems fine with this for some insane reason. Google isn't trying to be worse than Apple, they're trying to match Apple.
It's also worth noting that Microsoft has been doing the same thing just in a more mob like "shame if we and our antivirus partners would defame you" kind of way instead of outright disallowing unsigned programs for now.
Google are trying to match Apple's worst behaviour.
Next will be USA deciding that it's illegal to use the new Chinese offerings that will pop up to work around Google/Apple shittiness, because unconstitutional market manipulation goes hand-in-hand with USA's Russian-style oligarchy.
How long before Linux is added to the USA regime list of software that 'shows someone is a terrorist'.
>Corporation like Google should have been fined to the edge of bankruptcy for such predatory behaviour.
Google has an alternative at least (can download other app stores, apks, and have been able to for 15 years), I have no idea how Apple has been allowed to get away with what they do.
Better framing in their lawsuits- being a ‘market’ is less a thing in a monoculture- i.e. the market for Apple phones vs the market for all Android phones from a variety of makers.
Also in the worldwide market it most definitely isn’t a monopoly in phones at all- which is exactly why the EU had to reform their laws (DMA ect)to get regulatory powers over them.
"It's not a monopoly" is exactly the framing. Antitrust isn't inherently concerned with monopolies, it's concerned with market power. It's just that a monopolist has no room to move in claiming that they don't have market power.
They also play this trick where they pretend the only market is "phones" and ignore all of the complementary markets they're tying together with their phones, like app distribution and browser engines etc., since the market definitions for those shouldn't be the same as it is for "phones" when an app store that can be used on Google and Samsung phones can't be used (and therefore doesn't compete) on Apple phones and vice versa.
I never understood the argument that Apple's hostility to developers was unfair. The business proposition is simple:
"Buy our overpriced hardware with our famously inflexible and invasive software and sign this massive, legally-binding contract agreeing that you understand all of the above to extract further value out of the Apple ecosystem."
The "App Store" is just a natural extension of the club. As far as I'm concerned, they can gatekeep and sloppify as much as they want. Apple users agree to this.
I've never felt slighted that I couldn't hack apart my iPad - though I will absolutely do it once somebody cracks the code and gets a proper Linux kernel running.
Well they banned AI vocals at all too, so you can spend 6 months composing a track, add AI vocals, and your entire song is banned. Really the entire genre of edm could often switch out whatever vocals they have...
I have zero idea how much this matters since I don't really think anyone views the charts to find their next song to listen to. If a song with AI vocals is good it (might) pop off, I don't think getting on the charts at that point would matter.
This is like a pen manufacturer not wanting their pens used to sign drone strike orders, now the military needs to have a special box of pens that don't have stipulations attached. With AI usage it would be the same thing except applied to entire product chains. It seems like it would just add more complexity to operations.
You can agree with the rules anthropic wanted, but having rules set by a private company at all that apply to the military does seem fair for the military to object to.
>The Department reasonably feared that Anthropic might manipulate Claude’s design to prevent it from performing national-security functions that the Department deems contractually authorized and necessary
Though they'd probably put the DoD on the cybersecurity whitelist today, the very idea of the claude whitelists for certain functionality already exists and is being used by them today.
reply