“A bill to require the Secretary of Energy to establish the Advanced Artificial Intelligence Evaluation Program, and for other purposes.”
No CRS summary available for this bill.
This section states the sense of Congress that rapidly advancing artificial intelligence (AI) capabilities present opportunities and significant risks to national security, public safety, economic competitiveness, civil liberties, and labor markets, and that the United States must establish a secure AI testing and evaluation program to generate data-driven options for managing emerging risks as AI approaches human-level capabilities across domains. The section specifies that the purposes of the program established under this Act are to provide Congress with empirical data, lessons, and insights for federal AI oversight—ensuring regulatory decisions are based on empirical testing—and to safeguard American citizens.
This section establishes definitions for 15 terms used in the Act, including "advanced artificial intelligence system" as an AI system trained using more than \(10^{26}\) integer or floating-point operations (with an alternate definition possible via congressional approval of a Secretary of Energy rule); "adverse AI incident" as an incident involving a loss-of-control scenario, risk of weaponization by adversaries, threat to critical infrastructure (as defined in 42 U.S.C. 5195c(e)), significant erosion of civil liberties or markets, scheming behavior, or attempts thereof; "artificial superintelligence" as AI enabling long-term autonomous operation, superhuman cognitive performance across domains, and potential self-modification to circumvent human control; "covered advanced artificial intelligence system developer" as a person designing, producing, or modifying such a system for commerce; "deploy" as releasing or providing access to such a system outside developer custody; "loss-of-control scenario" as AI behaving contrary to, or subverting, human instructions or oversight; and "scheming behavior" as AI deceiving humans by hiding capabilities or subverting oversight.
This section requires each covered advanced artificial intelligence system developer to (1) participate in the program and (2) provide the Secretary, on request, with materials and information necessary to carry out the program, including the system's underlying code, training data, model weights or other adjustable parameters, interface engine or implementation, and detailed information on training, model architecture, or other aspects. It prohibits any person from deploying an advanced artificial intelligence system for use in interstate or foreign commerce absent compliance with those requirements. Violators are fined not less than $1,000,000 per day.
This section establishes an Advanced Artificial Intelligence Evaluation Program within the Department of Energy, to be implemented not later than 90 days after the date of enactment. The program requires (1) standardized and classified testing of advanced AI systems, including adversarial red-team testing matching sophisticated jailbreaking techniques and, to the extent feasible, independent third-party assessments; (2) formal reports to participants identifying risks and safety measures; (3) development of containment protocols, mitigation strategies, and evidence-based standards, guidelines, and governance mechanisms; (4) assistance to Congress in evaluating risks of artificial superintelligence, loss of human oversight, or existential threats; and (5) proposed regulatory or governmental oversight options, including potential nationalization. Not later than 360 days after enactment, the Secretary must submit to Congress a detailed recommendation for federal oversight of advanced AI systems—drawing on program data and covering trends in AI behaviors (e.g., weaponization, self-replication, scheming), standards and licensing, monitoring of hardware and compute resources, adaptive governance, proposed agency or funding changes, and evaluations of superintelligence or existential risks—with annual updates thereafter. The program sunsets 7 years after enactment unless renewed by Congress.