The Great American AI Act of 2026: Evidence Pack on Congress's Frontier Model Play

For the past two years, the regulation of artificial intelligence in the United States has been a fragmented, state-by-state experiment. That era may be ending. On June 4, 2026, Representatives Jay Obernolte (R-CA) and Lori Trahan (D-MA) released a 269-page bipartisan discussion draft titled the "Great American Artificial Intelligence Act of 2026" (GAAIA).[1][3]

The draft legislation represents Congress's most ambitious attempt to establish a comprehensive federal governance regime for AI. It arrives at a critical inflection point: states like California, Colorado, and Texas have been aggressively passing their own AI statutes, while the White House recently issued a June 2 Executive Order demanding early government access to next-generation models.[2][6]

Because the GAAIA is currently a "discussion draft"—released specifically to solicit feedback from stakeholders before formal introduction—its provisions are not yet law. However, legal analysts note that its framework signals a bipartisan consensus on how the federal government intends to regulate the technology's most powerful iterations.[3][7]

The central claim of the GAAIA is that the United States cannot maintain global leadership or domestic safety through a 50-state patchwork of regulations. To address this, Title I of the bill introduces a sweeping federal preemption clause. If enacted, the bill would freeze all state and local laws "specifically regulating the development of" any AI model for a period of three years.[1][3][5]

The evidence supporting this preemption strategy points to the growing compliance burden on developers. In May 2026, Colorado was forced to repeal and replace its landmark AI Act before it even took effect, illustrating the volatility of state-level rulemaking. By preempting state laws, the GAAIA aims to give the federal government exclusive jurisdiction over how models are trained and tested.[2][5]

However, there is significant uncertainty regarding the boundaries of this preemption. The draft explicitly preserves state laws of "general applicability," meaning existing consumer protection and privacy frameworks—like the California Consumer Privacy Act (CCPA)—would remain fully enforceable. Furthermore, the preemption only applies to the "development" of AI, leaving states free to regulate how AI is "deployed" or used by end consumers.[1][3]

The second major claim of the legislation is that regulatory scrutiny should be concentrated at the "frontier"—the layer where catastrophic risks originate. The bill creates binding federal obligations specifically for "large frontier developers." This category is defined by a strict financial threshold: companies with more than $500 million in annual revenue that have trained a frontier model.[1][7]

The evidence of these obligations is detailed in the draft's transparency and auditing requirements. Covered developers—which would include OpenAI, Anthropic, Google, and Meta—must publicly post risk assessments and mitigation steps. More significantly, they must submit to third-party audits conducted by designated Independent Verification Organizations (IVOs).[1][3]

The enforcement mechanisms attached to these frontier obligations are severe. The bill mandates that developers report "critical" safety incidents to federal authorities within 15 days. If an incident poses an "imminent risk of death or serious physical injury," the reporting window shrinks to 24 hours. Violations of these transparency requirements carry fines of up to $1 million per day.[3][7]

Beyond the technical governance of models, the GAAIA makes a third claim: that the federal government must actively monitor and mitigate AI's impact on human labor. Title II of the bill focuses entirely on workforce protections, establishing an AI Workforce Research Hub within the Department of Labor.[1][4]

The evidence for this workforce focus is found in the bill's data collection mandates. The Bureau of Labor Statistics and the Census Bureau would be required to update federal surveys to capture data on AI adoption and its correlation with mass layoffs. Additionally, the bill establishes robust federal whistleblower protections for employees and independent contractors who report "AI violations" at any company, not just frontier developers.[3][4]

The fourth claim addresses national security. Title III of the GAAIA focuses on cybersecurity, extending the Cybersecurity Information Sharing Act of 2015 through fiscal year 2035. This allows companies to share cyber threat information related to AI vulnerabilities without incurring antitrust liability.[3][5]

This cybersecurity focus aligns closely with the White House's June 2, 2026, Executive Order. That order established a voluntary framework for developers to provide the government with access to new frontier models up to 30 days before public release, specifically to assess their capabilities for cyberattacks. The GAAIA attempts to codify parts of this security apparatus into permanent law.[6][7]

To support these standards, the bill formally authorizes the Center for AI Standards and Innovation (CAISI) within the Department of Commerce. The draft allocates $100 million annually from 2027 through 2029 for CAISI to develop voluntary guidelines, evaluate AI systems, and monitor the progress of foreign competitors.[3][7]

Despite the comprehensive nature of the draft, the path to enactment remains highly uncertain. The preemption clause has already drawn fierce opposition from state-rights advocates and state Attorneys General, who argue that a three-year freeze on state lawmaking leaves consumers vulnerable if federal agencies move too slowly.[5][7]

Furthermore, labor advocates and civil society groups have criticized the bill's narrow focus on "frontier" models, arguing that it fails to address the immediate harms caused by smaller, deployed AI systems used in hiring, housing, and healthcare. As Congress heads toward its August recess, the discussion draft will face intense lobbying from both the tech industry seeking uniformity and consumer groups demanding stricter deployment safeguards.[4][5][7]