Advanced AI Security Readiness Act

NSA would create AI security playbook

If enacted, the National Security Agency would create an AI Security Playbook to protect advanced AI from theft and espionage. It would map weaknesses in AI data centers and developers, flag sensitive items like model weights and core methods, and set ways to detect, prevent, and respond to cyber threats. It would explain when security needs would require major U.S. Government involvement and sketch a highly secure government-run build, covering cybersecurity, weight protections, insider-threat vetting and clearances, access controls, counterintelligence, and emergency plans. The Director would review industry documents, interview experts, host roundtables, visit AI facilities, and work with a federally funded research center; these talks would not create a federal advisory committee. Reports would be due within 90 days (progress) and 270 days (final), with a public unclassified version for private-sector use and an optional classified annex; this would not grant new regulatory or enforcement powers.