Study on geopolitical strategies and verification frameworks for advanced artificial intelligence

(a) Strategy The Assistant Secretary for Emerging Threats, in coordination with the Assistant Secretary for Cyberspace and Digital Policy, shall develop a strategy to preserve enduring United States leadership in the field of advanced artificial intelligence (hereinafter in this section referred to as AI). (b) Matters To be included The strategy required by subsection (a) shall include the following: (1) A description and assessment of the technical and non-technical methods to monitor the present or future development or deployment of covered AI systems by adversarial nations or foreign actors, including China, with particular attention to systems that could pose severe threats to United States national security; (2) Proposals for potential commitments or agreements under which one or more foreign states would commit to restrict, limit, or halt the development or deployment of covered AI systems; (3) The identification and evaluation of monitoring, verification, and enforcement mechanisms, including methods that do not yet exist but could be developed or strengthened through additional research and development, that could be used to assess international compliance with the commitments or agreements described in paragraph (2) or otherwise improve United States national security, including hardware-based safeguards, data center inspections, cloud service audits, satellite monitoring, signals intelligence, and other relevant methods; (4) Identification and assessment of potential evasion techniques or deception strategies that adversaries could employ to circumvent verification mechanisms, and evaluate countermeasures to enhance the credibility and robustness of such mechanisms; (5) An examination of potential diplomatic and geopolitical strategies or options available to the United States in the event that the United States chose to delay the development or deployment of covered AI systems of adversaries, including options or plans to coordinate with allies and partners, build international confidence in potential agreements or commitments related to covered AI systems, deter opportunistic behavior by foreign adversaries, and support efforts to detect or verify that foreign adversaries were not secretly developing covered AI systems or attempting to covertly usurp United States leadership in advanced AI. This shall include methods, strategies, policies, and plans that, in either the presence or absence of international agreements, could be used to prevent or deter adversaries from developing, acquiring, or deploying covered AI systems that could pose severe threats to United States national security; and (6) The identification of potential limitations in the strategies, agreements, proposals, and mechanisms outlined in paragraphs (1) through (5), and develop policy recommendations to address such limitations, including via coordination with allies and partners. Such recommendations shall emphasize concrete, near-term diplomatic and policy actions the United States Government can take to build strategic leverage and preserve policy optionality as AI capabilities advance, including options to strengthen coordination with allies and partners, develop international agreements and commitments around covered AI systems, develop robust verification mechanisms for such agreements, improve situational awareness around foreign covered AI systems for the United States government and allies and partners, verify the location of adversaries’ high-end compute clusters, and coordinate with allies and partners to identify adversaries’ attempts to develop, acquire, or deploy covered AI systems. Such recommendations shall also include a prioritized list identifying verification methods or mechanisms that would benefit from additional research and development efforts by the United States, the private sector, public-private partnerships, or other relevant entities, with a focus on methods or mechanisms that could be made technically feasible or substantially enhanced within five years of additional research and development efforts. (c) Report Not later than two years after the date of the enactment of this Act, the Assistant Secretary for Cyberspace and Digital Diplomacy shall submit to the appropriate congressional committees a report that contains the strategy required by subsection (a). (d) Congressional notification During the 2-year period described in subsection (c), the Assistant Secretary for Cyberspace and Digital Policy shall provide timely updates to the appropriate congressional committees in response to significant developments related to the content of the report required by subsection (c), or other material developments in the global landscape of advanced artificial intelligence that may affect United States national security interests, verification strategies, or geopolitical stability. (e) Definition In this section, the term covered AI systems means advanced artificial intelligence (AI) systems with critical capabilities that the Secretary determines would pose a grave national security threat if developed, acquired, deployed, or stolen by adversaries, including systems that match or exceed human expert performance in high-risk domains such as chemical, biological, radiological, and nuclear matters, cyber offense, persuasion, or AI research and development, as well as systems that operate with high levels of autonomy without meaningful human oversight or that possess the capacity for self-improvement or rapid capability amplification.