Sam Altman says OpenAI shares Anthropic's red lines in Pentagon fight

Matrice Jacobine

OpenAI CEO Sam Altman wrote in a memo to staff that he will draw the same red lines that sparked a high-stakes fight between rival Anthropic and the Pentagon: no AI for mass surveillance or autonomous lethal weapons.
Why it matters: If other leading firms like Google follow suit, this could massively complicate the Pentagon's efforts to replace Anthropic's Claude, which was the first model integrated into the military's most sensitive work.
It would also be the first time the nation's top AI leaders have taken a collective stand about how the U.S. government can and can't use their technology.
The flipside: Altman made clear he still wants to strike a deal with the Pentagon that would allow ChatGPT to be used for sensitive military contexts.
Despite the show of solidarity, such a deal could see OpenAI replace Anthropic if the Pentagon follows through with its plan to declare the latter a "supply chain risk."
What he's saying: "[R]egardless of how we got here, this is no longer just an issue between Anthropic and the [Pentagon]; this is an issue for the whole industry and it is important to clarify our stance," Altman wrote Thursday evening in a memo obtained by Axios.
"We have long believed that AI should not be used for mass surveillance or autonomous lethal weapons, and that humans should remain in the loop for high-stakes automated decisions. These are our main red lines."
The intrigue: ChatGPT is already available in the military's unclassified systems, and talks to move it into the classified space have accelerated amid the Pentagon-Anthropic fight, sources tell Axios.
But the Pentagon has insisted OpenAI and Google would have to agree the military can use their models for "all lawful purposes," the same standard Anthropic rejected since it didn't incorporate their specific guardrails.
Elon Musk's xAI recently agreed to those terms, but Grok is not seen as a wholesale alternative to Claude.
In his memo, Altman wrote that the military will need AI, and he hopes to "help de-escalate things."
"We are going to see if there is a deal with the [Pentagon] that allows our models to be deployed in classified environments and that fits with our principles. We would ask for the contract to cover any use except those which are unlawful or unsuited to cloud deployments, such as domestic surveillance and autonomous offensive weapons," Altman said.
The Wall Street Journal first reported on the memo.
Between the lines: OpenAI's ideas for enforcing its red lines include preserving the company's ability to continuously strengthen its security and monitoring systems as it learns from real-world deployments, a source familiar told Axios.
The company also wants researchers with security clearances who can track how the technology is being used and advise the government on risks.
Finally, the source said, OpenAI wants certain technical safeguards — including confining models to the cloud rather than edge environments like autonomous weapons.
What to watch: Based on how Pentagon officials have described their position to Axios, those proposals could face the same resistance Anthropic encountered: too much private company influence over critical government work.
State of play: After Anthropic CEO Dario Amodei stood firm by his company's red lines, employees from OpenAI and Google signed onto a letter in solidarity on Thursday, pushing executives at their respective companies to resist "pressure" from the Pentagon.
While Anthropic said it intended to continue negotiations, a rupture appeared close. Emil Michael, the Pentagon official handling negotiations with Anthropic and the other major AI firms, denounced Amodei as a "liar" with a "God complex" who was "putting our nation's safety at risk."
Many others in D.C. and Silicon Valley praised Anthropic for taking a principled stand at the risk of a major financial hit.
Altman and Amodei are former colleagues at OpenAI who have become fierce rivals since the latter left to start Anthropic.
The other side: Defense officials contend they have no intention of conducting mass surveillance or swiftly deploying autonomous weapons.
Their primary objection is having a private company dictate how the U.S. government can deploy AI for national security purposes, particularly during a technological race with China.
Defense officials told Axios their interactions with Anthropic left them concerned the company might raise questions about the deployment of their technology at critical junctures. Anthropic denies that.
It's possible the negotiations with OpenAI will be less adversarial.
What to watch: "We have had some meetings to discuss this over the past couple of days, and will have more tomorrow with our safety teams before we decide what to do. We will also set up an all hands and office hours as soon as we can," Altman said, referring to those negotiations.
"This is a case where it's important to me that we do the right thing, not the easy thing that looks strong but is disingenuous. But I realize it may not 'look good' for us in the short term, and that there is a lot of nuance and context."

"We are going to see if there is a deal with the [Pentagon] that allows our models to be deployed in classified environments and that fits with our principles. We would ask for the contract to cover any use except those which are unlawful or unsuited to cloud deployments, such as domestic surveillance and autonomous offensive weapons," Altman said.

I'm happy to see this! I'm a bit suspicious that it's not the same red line Anthropic is taking, and that it's instead something that seems superficially similar but is optimized to get the administration to choose OpenAI over Anthropic. Hopefully time will tell. If OpenAI presents the same terms to the government as Anthropic, in solidarity, that would be a small but meaningful positive update on OpenAI for me. I currently expect them to somehow come out on top after the dust settles, emerging with some of the juicy government contracts that Anthropic loses out on.

I'm a bit suspicious that it's not the same red line Anthropic is taking, and that it's instead something that seems superficially similar but is optimized to get the administration to choose OpenAI over Anthropic

Yeah, this bit specifically:

We would ask for the contract to cover any use except those which are unlawful or unsuited to cloud deployments, such as domestic surveillance and autonomous offensive weapons

This seems to match how the DoW itself is framing the situation, which is that Anthropic's "red lines" are already covered by "only lawful use". So if OpenAI signs a DoW contract without any additional restrictions, precisely the way the DoW wants, this would still be technically in line with Altman's statement here.

But per Anthropic's refusal, there's a difference between the kinds of mass domestic surveillance that are illegal and the kinds of mass domestic surveillance that Anthropic wants to rule out:

To the extent that such surveillance is currently legal, this is only because the law has not yet caught up with the rapidly growing capabilities of AI. For example, under current law, the government can purchase detailed records of Americans’ movements, web browsing, and associations from public sources without obtaining a warrant, a practice the Intelligence Community has acknowledged raises privacy concerns and that has generated bipartisan opposition in Congress. Powerful AI makes it possible to assemble this scattered, individually innocuous data into a comprehensive picture of any person’s life—automatically and at massive scale.

So my current read is that Altman has once again managed to commit to exactly nothing.

Edit: Well, I guess he is still at least making a show of public support, which should increase the pressure on the DoW and at least make it look like the whole industry is against them? Unless this statement is intended to be parsed by the DoW as OpenAI offering them a line of retreat, a way to give them everything concrete they want while de-escalating the public narrative? Hm.

I'm confused - I parsed Altman's statement as:

We would ask for the contract not to cover use that is EITHER:
Unlawful
Unsuited to cloud deployments. Including:
Domestic surveillance
Autonomous offensive weapons

Isn't this equivalent to what Anthropic wants?

My parsing is:

We would ask for the contract not to cover use that is unlawful. Examples of unlawful use: domestic surveillance, autonomous offensive weapons.

As in, surveillance/autonomous weapons are framed as instances of unlawful use ("such as"). If we accept that framing, then any contractual language that merely says "no unlawful use" would technically cover them, even if it doesn't mention those items explicitly. The issue is that the law may be forbidding "domestic surveillance" and "autonomous weapons" under definitions of those terms that are importantly narrower than the custom (?) definitions on which Anthropic is insisting.

(Though I don't know that digging this deeply into Altman's Exact Words here makes sense. This may have just been a throwaway statement, without this heavy wording-optimization.)

Surely he meant something by "unsuited to cloud deployments"?

I wasn't sure what this meant, but I asked Claude, and it sounds like this is probably a reference to confidential military operations that require airgapping. This seems like a technical issue that has basically nothing to do with ethics, suggesting that the "ethical" part routes entirely through the "no unlawful use" part.

The part where Altman says "AI should not be used for mass surveillance or autonomous lethal weapons [...] These are our main red lines" sounds pretty unambiguous, but maybe this is easier to wriggle out of than a concrete statement about the contract OpenAI will request.

I don't see why it would be difficult to get data about movement patterns of individual people brought from data brokers into an airgapped environment.

The issue is that LLMs that run on the cloud can't interact with airgapped data.

Why? You put the data on a hard drive and then put the hard drive into your airgapped environment.

Even if we leave out the data brought from data brokers, the US military has plenty of spy satellites. Given the permission for doing foreign surveillance, there's a good chance that they are currently running agents in the Palantir system that analyze satellite imagery (and other sensor information) for foreign surveillance.

Given that a lot of those surveillance satellites aren't geostationary, they are likely already flying over the US and that data is available for analysis already.

Yep, on my read no supposed "redlines" are not actually in the contract language they have shared, e.g. consider whether this part in fact names a "redline":

https://x.com/i/status/2027846481021980914

I had thought that OpenAI, xAI, and GDM might leave Anthropic to face the music on their own. Then we got word of a petition signed by numerous GDM staff and about a dozen OpenAI staff, urging their employers to draw the same red lines as Anthropic, and I wondered if this was actually about xAI and Trump 2.0 against everyone else. Now we hear that xAI has in fact already agreed to everything the Pentagon wants, but Grok is not considered as capable or reliable as the other AIs.

I'm not sure how it will all turn out. But it reminds me most of when Trump 2.0 started putting pressure on the elite universities via their funding. The administration had turned its attention to a particular class of powerful institutions, made public demands of them all, and they each responded in their own way - Harvard resisted, Columbia folded. Now it's the turn of the frontier AI companies to be confronted.

In my opinion, the significance extends beyond internal US affairs and even beyond conflicts of the moment, such as a possible war with Iran, to the whole nature of the "Pax Silica" grouping which seems to be the first truly AI-driven geopolitical association, since these four companies are at the hub of it all, and what's being determined is their relationship to US hard power.

Now that war with Iran has actually started, I feel that this must have also been a factor in how Trump 2.0 tackled the situation with Anthropic. If Anthropic really did express misgivings about the use of Claude in the capture of Venezuela's Maduro, the Pentagon must have been concerned about what Anthropic could get up to during a war with Iran, a far larger operation.

I have no understanding of where and how frontier AI is playing a role in this war, but I wonder if any agency involved has already switched from Claude to "WarGPT"? I have yet to use an interface like OpenClaw, but I would guess that it's pretty easy to change the model you're using. Maybe it's only hard to change over, if you're using a finetuned model like a custom GPT.

If your have an agent that's works, you wouldn't want to switch it's model and get potentially more bugs right in the moment you start a war. The alternative models like the open sourced models from Meta that they have are likely worse than the frontier models they access from Anthropic and there might be performance degradation that you don't want to discover the moment you are waging war.

I would expect that the generals who are actually using the technology don't like it when they are told from above that they have to now start the 6-month process of moving off them when they would rather focus on the war.