Anthropic rejects Pentagon AI contracts over ethical concerns about surveillance and weapons

Anthropic has taken a definitive ethical stand against developing AI for certain military applications, refusing U.S. Department of Defense contracts related to mass domestic surveillance and fully autonomous weapons. In a statement from CEO Dario Amodei, the company argued that these uses could undermine democratic values and exceed the safe, reliable capabilities of current technology. This principled refusal has triggered a significant government response, with federal agencies moving to designate Anthropic a supply-chain risk and cease collaboration. The decision creates a stark contrast with competitor OpenAI, which simultaneously announced a new agreement to provide its AI models for use in classified Pentagon settings, a status Anthropic previously held.

The core of Anthropic's argument is that AI-powered mass surveillance, by assembling scattered data into comprehensive life profiles, presents novel risks to fundamental liberties that existing law has not yet addressed. Regarding autonomous weapons, the company draws a line between partially autonomous systems (which it supports) and fully autonomous ones that remove humans from the decision loop for target selection and engagement. This public rift highlights a growing schism in the AI industry between companies willing to engage broadly with defense and intelligence agencies and those establishing strict ethical red lines. The immediate consequence is a reshuffling of the Pentagon's AI vendor landscape, with OpenAI gaining ground as Anthropic cedes its position, setting a precedent for how AI firms navigate the complex intersection of technology, ethics, and national security.