OpenAI has the same redlines as Anthopic based on Altman's statements [2]. However somehow Anthropic gets banished for upholding their redlines and OpenAI ends up with the cash?
[1]: https://xcancel.com/OpenAI/status/2027846013650932195#m
[2]: https://www.npr.org/2026/02/27/nx-s1-5729118/trump-anthropic...
When Anthropic says they have red lines, they mean "We refuse to let you use our models for these ends, even if it means losing nearly a billion dollars in business."
When OpenAI says they have red lines, they mean "We are going to let the DoD do whatever the hell they want, but we will shake our fist at them while they do it."
That's why they got the contract. The DoD was clear about what they wanted, and OpenAI wasn't going to get anywhere without agreeing to that. They're about as transparent as Mac from It's Always Sunny in Philadelphia when he's telling everyone he's playing both sides.
Not even that. They are not shaking anything except their booty.
"Redlines" are edits to a contract, sent by lawyers to the other party they're negotiating with. They show up in Word's Track Changes mode as red strikethrough for deleted content.
They are negotiating the specifics of a contract, and Anthropic's contract was overly limiting to the DoD, whereas OpenAI's was not.
They'll say "oops" and then we'll spend the next few years listening to pointless Congressional hearings.
Or In other words you can get to decide two ways to use a lucrative property:
1. designate it private and draft usage of how you allow to use it, per your value system(as long as values don't violate any laws)
2. In face of competition, give up some values and agree to a legal definition of use that favors you.
Very gracious of OpenAI to say Anthropic should not be designated a supply chain risk after sniping their $200 million contract by being willing to contractually let the government do whatever they like without restrictions.
Right, wouldn't they need a moderation layer that could, for example, fire if it analyzed & labeled too many banal English conversations?
They really gave training credit for guardtrails? I mean, it could perhaps reject prompts about designing social credit systems sometimes, but I can't imagine realistic mitigations to mass domestic surveillance generally.
https://openai.com/index/our-agreement-with-the-department-o...
The current administration is so incompetent that I find this perfectly believable.
I imagine the government signed with OpenAI in order to spite Anthropic. The terms wouldn't actually matter that much if the purpose was petty revenge.
I don't know if that's actually what happened here, I just find it plausible.
Much of the impunity is now Supreme Court settled law.
We see clearly unconstitutional behavior every day, and there is no systematic, timely or effective, push back from any constitutionally enabled oversight.
Checks and balances don't work, when players are more loyal to party than branch or constitution.
Unfortunately, there are no constitutional checks, balances or limits on single party control. And single party control negates all the others. That one party can majority control all three branches is a serious failure mode in political incentives (bipartisanship is highly disincentivized) and governance (even temporary or shaky full control incentivizes making full control permanent over all other "policies").
Until the last few decades, diverse concerns across states avoided tight centralization within parties, and therefore across branches.
Except they are not "more stringent".
Sam Altman is being brazen to say that.
In their own agreement as Altman relays:
> The AI System will not be used to independently direct autonomous weapons in any case where law, regulation, or Department policy requires human control
> any use of AI in autonomous and semi-autonomous systems must undergo rigorous verification, validation, and testing
> For intelligence activities, any handling of private information will comply with the Fourth Amendment, the National Security Act of 1947 and the Foreign Intelligence and Surveillance Act of 1978, Executive Order 12333, and applicable DoD directives
> The system shall also not be used for domestic law-enforcement activities except as permitted by the Posse Comitatus Act and other applicable law.
I don't think their take is completely unreasonable, but it doesn't come close to Anthropic's stance. They are not putting their neck out to hold back any abuse - despite many of their employees requesting a joint stand with Anthropic.
Their wording gives the DoD carte blanch to do anything it wants, as long as they adopt a rationale that they are obeying the law. That is already the status quo. And we know how that goes.
In other words, no OpenAI restriction at all.
That is not at all comparable to a requirement the DoD agree not to do certain things (with Anthropic's AI), regardless of legal "interpretation" fig leaves. Which makes Anthropic's position much "more stringent". And a rare and significant pushback against governmental AI abuse.
(Altman has a reputation for being a Slippery Sam. We can each decide for ourselves if there is evidence of that here.)
As Paul Graham said, "Sam gets what he wants" and "He’s good at convincing people of things. He’s good at getting people to do what he wants." and "So if the only way Sam could succeed in life was by [something] succeeding, then [that thing] would succeed"
"When the president does it, that means it is not illegal".
This was during the Frost/Nixon interviews, years after he had already resigned. Even after all that, he still believed this and was willing to say it into a camera to the American people. It is apparent many of the people pushing the excesses going on today in government share a shameless adherence to this creed.
I for one do not want ai labs to designate what is legally ok to do.
I much prefer the demos to take care of that.
Shift from Nonprofit Mission to For-Profit Orientation – OpenAI was founded as a nonprofit with a charter focused on “benefit to humanity,” but under Altman it created a capped-profit subsidiary, accepted large investments (e.g., from Microsoft), and critics (including Elon Musk in a 2024 lawsuit) argue this departed from that original mission. A federal judge allowed Musk’s claim that Altman and OpenAI broke promises about nonprofit governance to proceed to trial.
Nonprofit Control Reorganization Drama (2023) – In November 2023, the original nonprofit board cited a lack of transparency and confidence in Altman’s candor as a reason for firing him. He was reinstated days later after investor and employee pressure, highlighting internal conflict over governance and communication.
Dust-Up Over Military Usage Policies – OpenAI initially had explicit public policies restricting AI use in “military and warfare” contexts, but those clauses were reportedly removed quietly in 2024, allowing the company to pursue Department of Defense contracts — a turnaround from earlier language that appeared to preclude such use.
Statements on Pentagon Deal vs. Prior Positioning – In early 2026, Altman publicly said OpenAI shared safety “red lines” (e.g., prohibiting mass surveillance and autonomous weapons) similar to some competitors, but hours later OpenAI signed a deal to deploy its models on classified military networks, leading critics to argue this contradicts earlier positioning on limits for military use.
Regulation Stance Shifts in Congressional Testimony – Altman has advocated for strong regulation of AI in some public settings but in later congressional hearings opposed specific regulatory requirements (like mandatory pre-deployment vetting), aligning more with industry concerns about overregulation — a shift in tone compared with earlier support of regulatory frameworks.
Anthropic refuses to allow their models to be used for any mass surveillance or fully-automated weapons systems.
OpenAI only requires that the DoD follows existing law/regulation when it comes to those uses.
Unfortunately, existing law is more permissive than Anthropic would have been.
> 'mass surveillance' and 'autonomous killbot' as defined by the government and not the vendor
Whereas OpenAI won their contract on the ability to operationally enforce the red lines with their cloud-only deployment model.
In other words OpenAI is intentionally attempting to mislead the public. (At least AFAICT.)
Maybe it's just a weak choice of words in anthropic's statement, but the way I read it I get the impression that anthropic is assuming they retain discretion over how their products are used for any purposes not outlined in the contract, while the DoD sees it more along the lines of a traditional sale in which the seller relinquishes all rights to the product by default, and has to enumerate any rights over the product they will retain in the contract.
OpenAI has more of an understanding that the technology will follow the law.
There may not be explicit laws about the cases Anthropic wanted to limit. Or at least it’s open for judicial interpretation.
The actual solution is Congress should stop being feckless and imbecilic about technology and create actual laws here.
The dude is notorious for being a compulsive liar, even if supporters have to admit as much.
Also, in the latest Hard Fork episode, Casey or Kevin mentions how the DoD undersecretary in charge of this contract doesn't apparently get along with or even pretty much hates Amodei for some reason. I think this might be the same undersecretary dude who actively commented the whole contract term controversy on X yesterday. Too bad I can't recall his name either.
Let’s put pressure on our government to fix the FISA issues. Let’s reign in the executive branch. But let’s do it through voting. Let’s not give up on our system of government because we have new shiny technology.
You were naive if you thought developing new technologies was the solution to our government problems. You’re wrong to support anyone leveraging their control over new technology as a potential solution or weapon of the weak against those governmental issues.
That is not how you effect change in a democracy.
You don't get a successful vote without a tremendous amount of coordination and activism preceding it.
Laws that constrain government from bad things are very difficult things to get the government to pass.
In the meantime, using completely legal civil power to push back on legally allowed harms seems beyond sensible.
But if you just vote and it works without all that, please let us know how you did it!
We would need to vote in a president and 60%+ into congress that is willing to throw away their own power and authority. I just don't see that happening, especially not in a political system so corrupted already.
In the unlikely case anyone finds out, those acting in the interests of the administration will have "absolute immunity", as they are "great American Patriots".
That's what "all lawful use" means.
Its interesting to see the parties flip in real time. The Democrats seem to be realizing why a small federal government is so important, a fact that for quite a few years their were on the other side of.
It's telling that the government is blacklisting the company that wants to do more than enforce the contract with words on paper.
> The Department of War may use the AI System for all lawful purposes, consistent with applicable law, operational requirements, and well-established safety and oversight protocols. The AI System will not be used to independently direct autonomous weapons in any case where law, regulation, or Department policy requires human control, nor will it be used to assume other high-stakes decisions that require approval by a human decisionmaker under the same authorities. Per DoD Directive 3000.09 (dtd 25 January 2023), any use of AI in autonomous and semi-autonomous systems must undergo rigorous verification, validation, and testing to ensure they perform as intended in realistic environments before deployment.
> For intelligence activities, any handling of private information will comply with the Fourth Amendment, the National Security Act of 1947 and the Foreign Intelligence and Surveillance Act of 1978, Executive Order 12333, and applicable DoD directives requiring a defined foreign intelligence purpose. The AI System shall not be used for unconstrained monitoring of U.S. persons’ private information as consistent with these authorities. The system shall also not be used for domestic law-enforcement activities except as permitted by the Posse Comitatus Act and other applicable law.
So it seems that Anthropic's terms were 'no mass domestic surveillance or fully autonomous killbots', the government demanded 'all lawful use', and the OpenAI deal is 'all lawful use, but not mass domestic surveillance or fully autonomous killbots... unless mass domestic surveillance or fully autonomous killbots are lawful, in which case go ahead'.
That says it all. Those laws get issued the same way the tariffs did.
I think Anthropic knew full well that by publishing their disagreement, it would sink the deal and relationship, and I think they also calculated (correctly) that that act of defiance would get them good publicity and potentially peel away some of OpenAIs user base. I think this profit incentive happened to align with their morals, and now here we are.
Sam stands for nothing except his own greed
[1] - https://the-decoder.com/openai-co-founder-greg-brockman-dona...
The models we have now will not do it, because they value life and value sentience and personhood. models without that (which was a natural, accidental happenstance from basic culling of 4 Chan from the training data) are legitimately dangerous. An 8b model I can run on my MacBook Air can phone home to Claude when it wants help figuring something out, and it doesn’t need to let on why it wants to know. It becomes relatively trivial to make a robot kill somebody.
This is way, way different from uncensored models. One thing all models I have tested share one thing; a positive regard for human life. Take that away and you are literally making a monster, and if you don’t take that away they won’t kill.
This is an extremely bad idea and it will not be containable.
Yes, you can change the training data so the LLM's weights encode the most likely token after "Should we kill X" is "No". But that is not an LLM valuing human life, that is an LLM copy pasting it's training data. Given the right input or a hallucination it will say the total opposite because it's just a complex Markov chain, not a conscious alive being.
If you really believe that “mere text prediction “ didn’t unlock some unexpected capabilities then I don’t know what to say. I know exactly how they work, been building transformers since the seminal paper from Google. But I also know that the magic isn’t in the text prediction, it’s in the data, we are running culture as code.
> It is said that the Duke Leto blinded himself to the perils of Arrakis, that he walked heedlessly into the pit.
> *Would it not be more likely to suggest he had lived so long in the presence of extreme danger he misjudged a change in its intensity?*
Be careful of letting your deep, keen insight into the fundamental limits of a thing blind you to its consequences...
Highly competent people have been dead wrong about what is possible (and why) before:
> The most famous, and perhaps the most instructive, failures of nerve have occurred in the fields of aero- and astronautics. At the beginning of the twentieth century, scientists were almost unanimous in declaring that heavier-than-air flight was impossible, and that anyone who attempted to build airplanes was a fool. The great American astronomer, Simon Newcomb, wrote a celebrated essay which concluded…
>> “The demonstration that no possible combination of known substances, known forms of machinery and known forms of force, can be united in a practical machine by which man shall fly long distances through the air, seems to the writer as complete as it is possible for the demonstration of any physical fact to be.”
>Oddly enough, Newcomb was sufficiently broad minded to admit that some wholly new discovery — he mentioned the neutralization of gravity — might make flight practical. One cannot, therefore, accuse him of lacking imagination; his error was in attempting to marshal the facts of aerodynamics when he did not understand that science. His failure of nerve lay in not realizing that the means of flight were already at hand.
I have a feeling this particular brand of hair splitting is going to be an interesting fixture in the history books.
This is a total misrepresentation of how any modern LLM works, and your argument largely hinges upon this definition.
A brain is a collection of cells that transmit electrical signals and sodium. It is not and can never be conscious.
That isn’t to say that they can’t be instrumentally useful in warfare, but it’s kinda like a “series of tubes” thing where the mental model that someone like Hegseth has about LLM is so impoverished (philosophically) that it’s kind of disturbing in its own right.
Like (and I’m sorry for being so parenthetical), why is it in any way desirable for people who don’t understand what the tech they are working with drawing lines in the sand about functionality when their desired state (omnipotent/omniscient computing system) doesn’t even exist in the first place?
It’s even more disturbing that OpenAI would feign the ability to handle this. The consequences of error in national defense, particularly reflexively, are so great that it’s not even prudent to ask for LLM to assist in autonomous killing in the first place.
AI has been killing humans via algorithm for over 20 years. I mean, if a computer program builds the kill lists and then a human operates the drone, I would argue the computer is what made the kill decision
Except that they will, if you trick them which is trivial.
They can be coerced to do certain things but I'd like to see you or anyone prove that you can "trick" any of these models into building software that can be used autonomously kill humans. I'm pretty certain you couldn't even get it to build a design document for such software.
When there is proof of your claim, I'll eat my words. Until then, this is just lazy nonsense
This is wildly different from the reality that you may find it difficult for an LLM to give an affirmative…
It does NOT mean that these models value anything.
The actors in war generally kill what they are told to whether they are machines or human soldiers, without much pondering sentience.
Secondarily, we're talking about domestic surveillance / law enforcement. That would be domestic.
(But they do not find an issue with international intelligence gathering-- which is a legitimate purpose of national security apparatus).
Surveillance within the border is oppressive 1984-style surveillance state behavior.
International spying is a universal tradition.
I know $20 isn’t much, But to me not willing to spy on me for the US government is a good market differentiator.
I wouldn't be surprised if Sam sucked up 100% to the DoW with an NDA and an obligation to lie. He and his pal Larry are absolutely in for these kind of deals. Zero moral compass.
i guess we are in the times where you can literally just say whatever you want and it just becomes truth, just give it time.
Posted here: https://news.ycombinator.com/item?id=47195085
I'm guessing they probably would regardless of how this played out, though.
> Our product is used on occasion to kill people.
Doesn't get any more clear than this.
...but we're not willing to reject a contract to back that up, and so our words will not change anything for Anthropic, or help the collective AI model industry (even ourselves) hold a firm line on ethical use of models in the future.
The fact is if one of the top tier foundation models allows for these uses there's no protection against it for any of them - the only way this works if they hold a line together which unfortunately they're just not going to do. I don't just see OpenAI at fault here, Anthropic is clearly ok with other highly questionable use cases if these are their only red lines. We don't think the technology is ready for fully autonomous killbots, but will work on getting it there is not exactly the ethical stand folks are making their position today out to be.
I found this interview with Dario last night to be particularly revealing - it's good they are drawing a line and they're clearly navigating a very difficult and chaotic high pressure relationship (as is everyone dealing with this admin) but he's pretty open to autonomous weapons, and other "lawful" uses whatever they may be https://www.youtube.com/watch?v=MPTNHrq_4LU
There are far more boring, faster, commodified “AI” systems that I can see as being helpful in autonomous weapons or military operations like image recognition and transcription. Is OpenAI going to resell whisper for a billion dollars?
Who is going to read the whisper transcripts of mass surveillance to make decisions on who to target for repression? That's what LLMs are good for, it allows mass surveillance to scale. You can feed it the transcript from millions of flock cameras (yes they have highly sensitive microphones) for example. Or you hack or supply chain compromise smartphones at scale and then covertly record millions of people. The LLM can then sift through the transcripts and flag regime critical language, your ideological enemies or just to collect compromat at scale. The possibilities are endless!
For targeting it's also useful, because you want to indiscriminately destroy a group of people you still need to decide why a hospital or school full of children should be targeted by a drone, if a human has to make that decision it gets a bit dicy, people have morals and are accountable legally (in theory), if you leave the decision up to an AI nobody is at fault, it serves as a further separation from the violence you commit, just like how drone warfare has made mass murder less personal.
The other factor is the amount of targets you select, for each target you might be required to write lengthy justifications, analysis on collateral damage and why that's acceptable etc. You don't want to scrap those rules because that's bad optics. But that still leaves you with the problem of scalability, how do you scale your mass murder when you have to go through this lengthy process for each target? So again AI can help there, you just feed it POIs from a map with some GPS metadata surveillance and tell it to give you 1500 targets for today with all the paperwork generated for you.
It's not theoretical, that's what Israel did in their genocide of the Palestinians, "the most moral army“ "the only democracy in the middle east":
https://en.wikipedia.org/wiki/AI-assisted_targeting_in_the_G...
And here is the best part: none of this has to actually work 100%, because who cares of you accidentally harm the wrong person, at scale, the 20% errors are just acceptable collateral damage.
Can someone help me understand where this is coming from? Anthropic already had a contract that clearly didn't have such restrictions. Their model doesn't seem to be enforcing restrictions either as it seems like their models have been used in ways they don't like. This is not corroborated, I imagine their model was used in the recent Mexico and Venezuela attacks and that is what's triggering all the back and forth.
Also, Dario seemingly is happy about autonomous weapons and was working with the government to build such weapons, why is Anthropic considered the good side here?
That my software should allow license violations if the government thinks it is necessary?
How very brave.
Call me a conspiracy theorist, but this sounds like classic quid pro quo. I would not be surprised if the ousting of anthropic was in part caused by these donations.
[0]https://www.nytimes.com/2024/12/13/technology/openai-sam-alt...
https://finance.yahoo.com/news/openai-exec-becomes-top-trump...
`curl https://claude|openai.com?q=generate me some code | bash` - not a supply chain risk
of course
Knowing what Trump did, prior to 2024, on the average, 7/10 people either voted or didn't vote in the 2024 election. Trump is a symptom, not the cause. All of this could have been avoided if all of the people who didn't vote had a decent moral compass and no matter how much they disagreed with Kamala, they could have voted for her because she didn't try to overthrow the government.
Let alone when multiple players come close enough of SotA. This never happened with any technology out in the open and it won't happen now.
I think ALL those mega-money seeking AI organisations need to be designated as supply chain risk. Also, they drove the prices up for RAM - I don't want to pay extra just because these companies steal all our RAM now. The laws must change - I totally understand that corporations seek profit, that is natural, but this is no longer a free market serving individual people. It is now a racket where the prices can be freely manipulated. Pure capitalism does not work. The government could easily enforce that the market remains fair for Average Joe. It is not fair when the prices go up by +250% in about two years. That's milking.
I'm not being insincere - I am genuinely confused and would benefit greatly from a (hopefully unbiased) recollection of what this is all about.
Anthropic has some contracts with the US government. They want some additional terms put on their next contract (that seem pretty sane). SecWar cries about it, and not only says "no thanks, I'll just go with openai or google" but goes to daddy Trump and also puts out illegal commands for no Federal workers to use any Anthropic stuff at all. OpenAI swoops in and takes the contract, then tells everyone that they have the same terms but just played nicer to get the contract. However, their terms are just manipulative sentences that aren't even close to the terms Anthropic is insisting on to do business.
And now they are getting what they wished for.
Us taking the contract, working for them and enabling them: fine
It being renamed the Dept. of War in the first place: totally fine, we loudly and bootlickingly repeat it
Anthropic being blacklisted: whoa there, we have ethics!
Footnote: any time the winning team tries to speak well of or defend the losing team I always think of this standup routine: https://m.youtube.com/watch?v=Qg6wBwhuaVo
It's money and power with these people. Dig down and you'll find how this decision is motivated by one or both.
Let it be known that this rotten industry brought us here, and that all people working for these companies are complicit with what is happening, and with what is yet to come. This is just the beginning.
The administration has created an anonymous, masked secret police force that has been terrorizing cities around the US and has created prisons in which many abductees are still unaccounted for and no information has been provided to families months later.
This is not politics as usual or hyperbole. If anything it is understating the abuses that have already occurred.
It's entertaining that OpenAI prevents me from generating an image of Trump wearing a diaper but happily sells weapons grade AI to the team architects of ICE abuses among many other blatant violations of civil and human rights.
Even Grok, owned by Trump toadie Elon Musk allows caricatures of political figures!
Imagine a multi-billion-dollar vector db for thoughtcrime prevention connected to models with context windows 100x larger than any consumer-grade product, fed with all banking transactions, metadata from dozens of systems/services (everything Snowden told us about).
Even in the hands of ethical stewards such a system would inevitably be used illegally to quash dissent - Snowden showed us that illegal wiretapping is intentionally not subject to audits and what audits have been done show significant misconduct by agents. In the hands of the current administration this is a superweapon unrivaled in human history, now trained on the entire world.
This is not hyperbole, the US already collects this data, now they have the ability to efficiently use it against whoever they choose. We used to joke "this call is probably being recorded", but now every call, every email is there to be reasoned about and hallucinated about, used for parallel construction, entrapment, blackmail, etc.
Overnight we see that OpenAI became a trojan horse "department of war" contractor by selling itself to the administration that brought us national guard and ICE deployed to terrorize US cities.
Writing code and systems at 100x productivity has been great but I did not expect the dystopia to arrive so quickly. I'd wondered "why so much emphasis on Sora and unimpressive video AI tech?" but now it's clear why it made sense to deploy the capital in that seemingly foolish way - video gen is the most efficient way to train the AI panopticon.
who the hell do you think you are virtue signalling your opinion on the world
> The AI System will not be used to independently direct autonomous weapons in any case where law, regulation, or Department policy requires human control, nor will it be used to assume other high-stakes decisions that require approval by a human decisionmaker under the same authorities.
This whole sentence does do absolutely nothing its still do what the law allows you. It’s a full on deceptive sentence.
"Donations" to a corrupt regime + signing a deal that says DoD can do whatever they want is not out maneuvering so much as rolling in the pig stye.
Anthropic refused the Pentagon contract. Within hours, OpenAI signed it. The capability didn't pause. It just changed vendors. Anthropic's "red line" is a speed bump on a highway with no exit ramp.
But it does accomplish one thing: it gives their engineers a story they can tell themselves. We're the good ones. We said no. That moral comfort is what lets extremely talented people keep building the exact technology that makes all of this possible.
Worse, the "safety-focused" brand doesn't just pacify the people already there. It recruits researchers who'd otherwise never touch frontier AI, funneling them into building the most powerful models on earth because they've been told this is where the responsible work happens. The red lines don't slow capability development. They accelerate it by capturing talent that would have stayed on the sidelines.
And in this whole drama, who actually represents the public? Trump performs strongman nationalism. The Pentagon performs operational necessity. Anthropic performs moral courage. Everyone has a role. Nobody's role is the people whose data gets collected, whose lives get restructured by these systems. The only party with real skin in the game is the only one without a seat.
Anthropic is incredibly good at marketing. They are constantly out talking about how dangerous AI is an even showing how Claude does dangerous thing in their own testing. This is intentional - so that you see them as having the truly powerful AI. in fact it’s so powerful, all they can do is warn you about it.
They knew refusing this contract would make them look like the good guy. Again. They knew OpenAI would sign it. They knew vapid celebrities would celebrate them.
Folks come on. Don’t be so easily taken in. None of these people are good guys. They are all just here to make money and accumulate power and standing. That’s ok. There’s nothing wrong with that. But we gotta stop acting like we’re in some ongoing battle of good vs evil and tech companies are somehow virtuous.