Red Kraken: an AI adversary for a congressional wargame
I built Red Kraken, the adversary agent in a closed-door congressional wargame on AI-enabled cyber threats to U.S. critical infrastructure in a China-Taiwan crisis, and ran it live during play. The defensive side was played by members of the House Committee on Homeland Security and the House Select Committee on China, with Rep. Michael McCaul leading Member participation. It ran as a multi-turn, simultaneous-allocation exercise against cyber targets across freight rail, port terminals, transportation networks and domestic infrastructure.
The agent reasons from a corpus of Chinese military doctrine and more than 150 documented PRC cyber operations, with a separate intelligence agent to predict U.S. moves. Dynamic point allocation means every game is different, illuminating new insights and furthering readiness against foreign threats.
CSIS published the exercise on 28 July 2026. Its analysis reports what the agent's play showed: concentration on West Coast ports for 3-5 day disruptions, later game pressure on logistics software and IT services rather than on infrastructure directly, and a shift toward undefended AI-enabled systems when traditional targets were hardened.
The agent is an analytical and educational tool. It isn't connected to any operational system, doesn't identify vulnerabilities in real infrastructure, and doesn't predict specific attacks. All source material is unclassified and publicly available.
