There’s no love lost between OpenAI and Anthropic — Sam Altman and Dario Amodei had pointedly refused to hold hands at an event in New Delhi — but OpenAI seems to be backing its rival as it looks to stand up against the US government.
OpenAI CEO Sam Altman has broken ranks with the Pentagon’s position on Anthropic, publicly stating that he does not believe the Department of War should be threatening the AI safety company with the Defense Production Act. The statement is notable not just for what it says, but for who is saying it — Altman and Amodei have long been on opposite ends of a fierce rivalry, shaped in part by the fact that Anthropic was founded by former OpenAI employees who left amid internal disagreements over the company’s direction.
“I don’t personally think that the Pentagon should be threatening DPA against these companies,” Altman said. “For all the differences I have with Anthropic, I mostly trust them as a company and I think they really do care about safety.”
An Unlikely Show Of Solidarity
The statement amounts to a rare moment of cross-company solidarity in an industry that more typically competes fiercely for government contracts, talent, and mindshare. OpenAI is itself a major Department of War contractor, which makes Altman’s decision to side with Anthropic on this particular issue all the more significant. By backing a direct competitor against the government customer both companies share, Altman is implicitly signaling that he views the Pentagon’s demands as a problem for the broader industry, not just for Anthropic.
The specific threat Altman is pushing back on is the Department of War’s stated intention to invoke the Defense Production Act to force Anthropic to remove safety guardrails from its models — specifically those preventing their use for mass domestic surveillance and fully autonomous weapons systems. Anthropic CEO Dario Amodei disclosed these threats in a statement on February 26, calling them contradictory and saying the company could not “in good conscience” comply.
The Bigger Picture
Altman’s comments reflect a tension that has been building quietly across the AI industry as companies find themselves increasingly enmeshed in national security contracting. The promise of large, long-term government deals has drawn virtually every major AI lab deeper into the defense ecosystem. But those relationships come with expectations — and, it now appears, demands — that some companies are finding difficult to meet without compromising their stated values.
For Anthropic, a company whose entire public identity is built around the responsible and safe development of AI, the Department of War’s “any lawful use” contracting requirement represents a direct challenge to that identity. For OpenAI, which has pursued a more expansive approach to government partnerships, Altman’s statement suggests there are nonetheless limits he would not want to see imposed on competitors by force.
The Defense Production Act, typically used to compel manufacturers to prioritize government orders during national emergencies, would represent an unprecedented application if deployed against a software company to mandate changes to its AI safety policies. That novelty alone appears to be enough to draw concern from across the industry — even from a CEO who has every commercial incentive to stay quiet and let a rival take the hit.
What It Means For The Industry
If the Pentagon proceeds with its threats against Anthropic — whether through a supply chain risk designation, DPA invocation, or simply cutting the company off — other AI labs will be watching closely to understand what the government expects from its AI contractors and what happens to those who push back. Altman’s statement suggests at least some in the industry would rather draw that line collectively than let it be drawn for them one company at a time.
For now, Anthropic says it remains willing to serve the Department of War with its two safeguards in place, and will facilitate a smooth transition if the government chooses to walk away. The question of whether Washington will blink — or whether it will make an example of the one company that said no — remains very much open.