H.R. 9965 — AI Threat Output and Monitoring Incident Containment Act

This section establishes the Advanced Artificial Intelligence Nuclear Evaluation Program within 90 days to test and evaluate artificial intelligence for the likelihood of nuclear incidents, implement jailbreaking and red-teaming protocols, facilitate independent assessments, and develop containment and mitigation recommendations. It requires large advanced artificial intelligence developers to participate, prohibits false statements, obstruction, or failure to provide model access, authorizes civil penalties of up to $1 million per violation (with each day of a continuing violation counted separately), and exempts submitted information from disclosure under the Freedom of Information Act except in specified circumstances.