The Great American Artificial Intelligence Act, a draft bill announced by Representatives Jay Obernolte (R-CA) and Lori Trahan (D-MA), would create the first comprehensive federal AI governance system in the United States—but most of its major provisions would expire after just three years. The over 250-page proposal, analyzed in a June 17, 2026 report by the Cato Institute, establishes transparency mandates for large AI developers, preempts certain state laws, and creates a new federal center to evaluate AI safety claims. The bill's authors frame it as a way to replace the current "informal, ad hoc evaluation process" with a stable regulatory regime that gives companies clear rules while protecting public safety.
The bill targets "large frontier AI developers" earning over $500 million in yearly revenue, requiring them to submit two key documents for each model release: a frontier AI framework outlining how the lab approaches safety and catastrophic risk, and a transparency report detailing the model's capabilities, intended uses, and risk assessments. These documents must be audited by independent third-party organizations licensed by the government, with audits occurring twice a year starting one year after the bill passes. The proposed Center for AI Standards and Innovation would receive $100 million annually to conduct its own model evaluations, represent U.S. interests in international standard-setting forums, and oversee the licensing of independent auditors. The bill also extends cybersecurity information-sharing protections from 2025 to 2035, shielding companies from antitrust enforcement when they collaborate on cyber threats.
According to the Cato report's author Juan Londoño, the transparency regime "forces companies to 'show their work' and leave a paper trail" so regulators can determine after an incident whether harm resulted from negligence or unforeseeable developments. The bill's preemption provision would block state laws on AI development—but explicitly allows states to continue regulating AI deployment and use, creating what the report describes as a split similar to automobile regulation where "car manufacturing is handled at the federal level" while states control how cars operate on their roads. On workforce preparation, the report notes that the bill "seems to place a higher focus on retraining and guiding potentially displaced workers" rather than heavy subsidies or tax-driven mandates on business staffing decisions.
The bill's design aims to prevent situations like the one that occurred June 12, 2026, when the government issued an export control directive suspending access to Anthropic's Claude Fable 5 and Mythos 5 models—forcing the company to block all customers due to compliance complexity. By requiring labs to conduct their own safety assessments verified by independent auditors, the proposal tries to balance flexibility for companies with government oversight capability. The report argues this approach "can have pro-innovation effects by replacing unpredictable, ad hoc regimes that largely rely on regulators' discretion." The bill also directs the National Science Foundation to establish eight regional "Centers of AI Excellence" for community colleges and tasks the Department of Labor with improving data collection on AI's workforce impact through expert workshops and new monitoring tools.
The biggest question mark hangs over the three-year sunset clause attached to nearly all major provisions—including the new federal center, the transparency requirements, the independent auditor system, and state law preemption. The Cato report warns this "would subject this complex governance system to the whims of volatile political winds, potentially erasing any progress achieved during the three-year period if these provisions are not reauthorized." While Congress can extend the provisions, the report notes this raises serious doubts about whether a bill meant to establish long-term AI governance can succeed when its core infrastructure is designed to disappear unless lawmakers vote to save it. For now, the draft represents the most detailed federal attempt yet to regulate AI development—if it can survive its own expiration date.