Anthropic Tells DOD: We Can't Tamper With Claude Mid-War
Anthropic fires back at Pentagon claims it could manipulate Claude AI models after military deployment.
Anthropic is pushing back hard against Department of Defense accusations that the AI maker could meddle with its Claude models once they're in military hands. In a new filing, the company flatly denies the possibility.
The DOD alleged that Anthropic could manipulate deployed AI models during active wartime operations — a serious claim with massive national security implications. Anthropic executives say that's technically impossible. Once Claude is deployed within military infrastructure, the company argues it simply has no mechanism to reach in and alter the model's behavior.
The dispute highlights a growing tension between AI developers and the defense establishment over control, trust, and the mechanics of how large language models actually work once they leave the builder's hands. It's a question that will only get louder as military AI adoption accelerates.