This bill directs the Director of the National Security Agency, through the Artificial Intelligence Security Center, to create an "AI Security Playbook." The Playbook must identify vulnerabilities in advanced AI data centers and developers, list components or information (for example, models, model weights, architectures, and core algorithmic insights) that would greatly help an adversary, and set out strategies to detect, prevent, and respond to cyber threats targeting covered AI technologies. The Playbook must analyze when the United States would need substantial involvement in building or overseeing highly secure AI systems and describe how that involvement could work, including cybersecurity protocols, protections for model weights, insider-threat mitigation, access controls, counterintelligence measures, and contingency plans. The Playbook may include a classified annex and must also include an unclassified portion with general guidelines for dissemination. The Director must engage prominent AI developers and researchers, collaborate with a federally funded research and development center with relevant experience, and submit an initial report within 90 days and a final report within 270 days after enactment.
If you work on advanced AI systems, the Playbook will produce unclassified guidelines and best practices intended for private-sector use. The Director is required to consult with prominent AI developers and may visit facilities, hold interviews, roundtables, and expert panels. The bill identifies categories of "covered AI technologies" that could pose a grave national security threat if stolen (for example, systems that match or exceed expert human performance in areas like chemical, biological, radiological, and nuclear matters, cyber offense, or autonomous decision-making). The bill does not itself create regulatory requirements in the section that asks the Director to identify security levels that might need substantial government involvement.
No publicly available information on costs, appropriations, or funding levels is included in the bill text. The bill sets reporting deadlines but does not specify funding to carry out the requirements.
No publicly available information beyond the bill text. The bill text states the objective: to develop strategies to defend covered AI technologies from technology theft by threat actors and to produce both classified and unclassified guidance.
No publicly available information on opponents' views or formal objections is included in the bill text or provided metadata.