The U.S. Department of Defense is considering ending its relationship with artificial intelligence company Anthropic amid a growing dispute over how its AI models can be used in military contexts. According to reporting by Semafor and other outlets, the Pentagon threatened to terminate a roughly $200 million contract after Anthropic refused to allow unrestricted use of its AI tools, especially in areas like autonomous weapons and mass surveillance — uses the company wants to keep off-limits.
Pentagon officials are reportedly pushing leading AI developers — including OpenAI, Google, xAI, and Anthropic — to agree that their systems can be deployed by the military for “all lawful purposes,” which encompasses weapons development, intelligence gathering, and battlefield operations. Anthropic has resisted those broad terms, maintaining ethical safeguards on its Claude models to prevent use in fully autonomous weaponry or domestic surveillance without adequate human oversight.
The clash comes after reports that the U.S. military used Anthropic’s Claude in a classified operation to capture Venezuelan President Nicolás Maduro, a move that raised alarms within the AI company given its prohibitions on violent or surveillance-related uses. Anthropic has publicly stated it has not had discussions about specific operational deployments with the Pentagon, but the overall sticking point remains how far military access to its AI should go.
If the Pentagon follows through on cutting ties or designating Anthropic a supply-chain risk — a status usually reserved for hostile foreign entities — it could require defense partners to stop using Claude entirely, reshaping the landscape of AI suppliers in U.S. defense. The dispute highlights wider tensions between AI safety advocates and national security priorities as governments seek powerful AI tools while companies push back on ethical limits.
