Anthropic Walks Away From Pentagon Over Surveillance and Autonomous Weapons
Mar 1st 2026
Negotiations collapsed after the Pentagon sought looser limits on Anthropic’s AI use, especially for bulk domestic data and for autonomous weapon systems, prompting a government move to cut ties.
- The Pentagon tried to renegotiate Anthropic’s contract to remove some company-imposed ethical limits and included qualifier language that worried Anthropic.
- Anthropic rejected Pentagon plans to analyze bulk data about Americans, calling that demand a dealbreaker.
- Anthropic refused to allow its models to be embedded in weapons and said a cloud-only promise would not prevent effective weapon integration because cloud and edge systems blend.
- Pete Hegseth moved to terminate the government relationship and directed military contractors to stop doing business with Anthropic.
- Anthropic’s model is currently the only AI allowed into classified systems while OpenAI quickly reached a separate cloud-only deal with the Pentagon.
- Nearly 100 OpenAI employees signed a letter backing red lines on mass domestic surveillance and autonomous weapons, highlighting internal dissent.