OpenAI Wins Pentagon Classified-Systems Deal, Agreeing to "All Lawful Purposes" With Its Own Technical Safeguards

What happened

OpenAI announced Friday that it had reached an agreement with the Pentagon to provide its artificial intelligence technologies for classified systems, the Associated Press reported on October 16 (via the Press Democrat). The announcement came just hours after President Donald Trump ordered federal agencies to stop using AI technology made by rival Anthropic. OpenAI CEO Sam Altman confirmed the deal in a social media post the same day.

Under the deal, OpenAI agreed to let the Pentagon use its AI systems for "all lawful purposes." At the same time, Altman said the company had found a way to ensure its technologies would not be applied to domestic mass surveillance in the United States and that autonomous weapons use would remain subject to human accountability — by installing specific technical guardrails in its systems. OpenAI will also embed engineers to work alongside government personnel on classified projects, helping deploy the models and ensure their safety.

The deal is the direct aftermath of the breakdown in talks between Anthropic and the Pentagon. The two sides had tussled for weeks over a $200 million contract: the Pentagon demanded it be able to use Anthropic's AI for all lawful purposes or it would cut the company off from government business, while Anthropic insisted on terms ensuring its technology would not be used for domestic surveillance of Americans or autonomous lethal weapons. They failed to agree by a 5:01 p.m. Friday deadline, after which Defense Secretary Pete Hegseth designated Anthropic a "supply-chain risk to national security," and Trump called it a "radical Left AI company."

Altman began negotiating with the Pentagon on Wednesday, two people familiar with the discussions told the AP. In a twist, Altman had publicly backed Anthropic's position earlier the same week, telling CNBC on Friday that he mostly trusted Anthropic and that "they really do care about safety"; dozens of OpenAI employees signed an open letter this week urging AI companies to stand together so the Pentagon could not divide them. The AP also reported that the Pentagon reached an agreement earlier this week to use xAI's technology for classified operations, with Google in similar talks.

“In all of our interactions, the DoW displayed a deep respect for safety and a desire to partner to achieve the best possible outcome.”

— Sam Altman, CEO of OpenAI (social media post, as reported by the Associated Press)

Key facts

1. Classified-systems deal: On October 16, OpenAI announced an agreement with the Pentagon to provide AI technology for its classified networks
2. "All lawful purposes" plus safeguards: Agreed to all lawful purposes; installed its own technical safeguards barring US domestic mass surveillance, with autonomous weapons under human accountability
3. Embedded engineers: OpenAI staff will work alongside government personnel on classified projects to assist deployment and ensure model safety
4. Anthropic out: The sides missed the 5:01 p.m. Friday deadline on a $200 million contract; Hegseth then designated Anthropic a "national security supply-chain risk," cutting it off from US government business
5. Supplier reshuffle: The Pentagon reached a classified-systems agreement with xAI earlier this week; Google is also in talks