The Department of War designated Anthropic, maker of the Claude AI models, a supply-chain risk to national security, a label previously reserved for foreign companies like Huawei. The move capped weeks of tension over Anthropic’s refusal to grant the military full access to its technology. Meanwhile, OpenAI announced its own classified-network deal with the DOW, claiming similar red lines but under terms Anthropic had rejected as unenforceable.

Join us for a brown bag lunch to unpack the standoff between AI safety commitments and national security imperatives. How does the Department of War conceive of autonomous weapons, and where do frontier AI models actually fit into the kill chain? What role is AI already playing in active military operations, including in Iran? And what drove the breakdown between the DOW and one of its most deeply embedded AI partners?

Panelists:

Andrew Reddie Associate Research Professor at UC Berkeley’s Goldman School of Public Policy and founder of the Berkeley Risk and Security Lab. His research examines how emerging military technologies shape international security, with a focus on AI governance, nuclear weapons policy, and wargaming.

Sarah Shoker Postdoctoral researcher at UC Berkeley and former head of the Geopolitics Team at OpenAI, where she led research on the implications of artificial general intelligence for international security and global stability

Chris Hoofnagle Professor of Law in Residence at Berkeley, longtime advisor to Palantir Technologies, and author of “How the AIs Kill,” a forthcoming article on how machine learning is used in targeting.