A contract spat between the Department of Defense and AI company Anthropic has hardened into a values‑driven standoff over the military’s use of generative AI, culminating Friday in an announcement that the Pentagon will no longer use Anthropic’s Claude product while also prohibiting military contractors from using it.
At the core of the dispute are Anthropic’s refusal to allow its model to not be used for mass surveillance of Americans nor power fully autonomous weapons that can engage targets without human oversight. Defense officials, in turn, demand the ability to deploy Anthropic’s Claude system for “all lawful purposes,” arguing that vendors’ stipulations should not constrain battlefield decisions even if those uses never actually materialize. The Pentagon has denied plans to use AI for mass surveillance of US citizens or for autonomous weapons.
OpenAI, maker of ChatGPT and its underlying AI models, has said they will accept the Pentagon’s preferred “all lawful purposes” language while relying on internal policies to bar mass surveillance of Americans and fully autonomous weapons, allowing it to secure a major defense‑AI role without matching Anthropic’s contractual red lines.
