On Feb. 26, Anthropic, an AI safety and research company and developer of Claude AI, released a statement regarding the Department of Defense’s use of its AI technology. In July 2025, the company made a $200 million contract with the Pentagon for its use in military operations, with two key limitations:

• Their AI could not be used for mass domestic surveillance. Anthropic stated, “AI-driven mass surveillance presents serious, novel risks to our fundamental liberties.” There are concerns that AI technology could be too effective at connecting public information, which could be used to make a comprehensive profile of individuals regardless of criminal status. The company did give permission for their AI to be used for foreign and counterintelligence missions.
• Their AI could not be used to make fully autonomous weapons. Anthropic believes that current AI technology is not reliable enough to be trusted with the power to take lives. They want there to be human oversight and proper guardrails before this step is taken. They offered to work on R&D to improve reliability for this purpose, with no reply from the Department of Defense. Partially autonomous weapons are allowed in the contract.

At the time of the statement’s release, Claude had already been used in military operations, including use in classified government networks and the capture of Nicolás Maduro.

In response to Anthropic’s decision to stand its ground, President Donald Trump attacked the company on Truth Social on Feb. 27, calling the limitations an attempt to “STRONG-ARM the Department of War,” and announcing that use of the company’s technology would be completely phased out of US agencies. In addition, Defense Secretary Pete Hegseth designated Anthropic a “Supply-Chain Risk to National Security,” giving his department six months to transition to a new service. Hours later, OpenAI, developer of ChatGPT, announced a deal with the Pentagon to use its AI technology.

OpenAI released a statement on Feb. 28 outlining its own limitations for the Pentagon’s use of its AI technology:

“No use of OpenAI technology for mass domestic surveillance.”

“No use of OpenAI technology to direct autonomous weapons systems.”

“No use of OpenAI technology for high-stakes automated decisions (e.g. systems such as “social credit”).”

These limitations are very similar to those outlined by Anthropic. The company states it is relying on its “safety stack, the contract language, and existing laws” to hold the government to the above limitations. It is unclear how OpenAI was able to reach an agreement with more limitations than Anthropic.