OpenAI signs Pentagon AI deal with safety guardrails
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Sam Altman announced late Friday that OpenAI has reached an agreement with the Department of Defense—now styled the Department of War under the Trump administration—to deploy OpenAI's AI models on the department's classified network.
- Anthropic and the Pentagon deadlocked over "all lawful purposes" language, with CEO Dario Amodei saying his company "never raised objections to particular military operations" but drew red lines around mass domestic surveillance and fully autonomous weapons.
- More than 60 OpenAI employees and 300 Google employees signed an open letter this week asking their employers to back Anthropic's red lines, highlighting internal tension at OpenAI even as Altman closed the Pentagon deal.
- President Trump publicly attacked Anthropic as "Leftwing nut jobs" and ordered federal agencies to phase out the company's products over six months, while Secretary of Defense Pete Hegseth designated Anthropic a supply-chain risk, barring military contractors and suppliers from any commercial activity with the firm.
- Anthropic said it had "not yet received direct communication from the Department of War or the White House on the status of our negotiations" and pledged to "challenge any supply chain risk designation in court."
- Altman claimed OpenAI's deal explicitly prohibits domestic mass surveillance and requires human responsibility for use of force, including autonomous systems, saying the DoW "agrees with these principles, reflects them in law and policy, and we put them into our agreement."
- OpenAI will build its own "safety stack" of technical safeguards, deploy engineers alongside the Pentagon, and—per Altman at an all-hands meeting—will not be forced by the government to override a model's refusal of a task, while pushing the DoW to offer identical terms to all AI companies.
Why it matters: The Pentagon now has one major AI vendor it can work with (OpenAI) and one it has publicly blacklisted (Anthropic, via a supply-chain risk designation Anthropic says it will fight in court). The roughly 360 Google and OpenAI employees who signed the pro-Anthropic letter suggest the internal debate over military AI use is far from settled inside the companies doing the work.
