Two things define you: your patience when you have nothing and your attitude when you have everything.

Sam Altman Says OpenAI's Pentagon Deal Has the Safeguards Anthropic Wanted

What started as a disagreement over contract terms quickly turned into one of the most dramatic standoffs in AI policy this year. The Pentagon asked AI companies to allow their models to be used for "all lawful purposes." Anthropic pushed back, drawing a hard line against mass domestic surveillance and fully autonomous weapons. OpenAI just signed a deal with the Department of Defense instead.

Sam Altman announced the agreement on Friday, but with a notable twist. He says the deal includes the very safeguards Anthropic was fighting for. According to Altman, OpenAI's contract explicitly prohibits the use of its models for domestic mass surveillance and requires human oversight over any use of force, including autonomous weapon systems. The government also agreed that if an OpenAI model refuses to perform a task, it cannot be forced to comply. OpenAI will embed engineers with the Pentagon to monitor how the models are being used.

Anthropic did not get the same outcome. After the two sides failed to reach an agreement, President Trump publicly criticized the company on social media and directed federal agencies to begin phasing out Anthropic's products over six months. Defense Secretary Pete Hegseth went further, designating Anthropic as a supply chain risk and barring any military contractor or supplier from doing business with the company. Anthropic, for its part, says it has not received any official communication about the designation and has vowed to challenge it in court.

The situation has divided the broader AI industry. More than 60 OpenAI employees and 300 Google employees signed an open letter this week expressing support for Anthropic's original position. Altman responded by saying he hopes the terms OpenAI secured can be offered to all AI companies, framing it as a path toward de-escalation.

What this episode makes clear is that the question of how AI companies engage with government and military applications is no longer theoretical. It is now a commercial, legal, and political battleground, and different companies are drawing very different lines.