Great American AI Act: Bipartisan Bill Proposes Three-Year Freeze on State AI Laws

What Just Happened

On June 4, 2026, Representatives Jay Obernolte (R-CA) and Lori Trahan (D-MA) released the discussion draft of the Great American AI Act, a 269-page bipartisan bill that would establish the first comprehensive federal framework for governing artificial intelligence in the United States. The draft was made public for stakeholder review and comment, with feedback requested at GAAIA@mail.house.gov before the legislation is formally introduced.

The bill's arrival comes less than a week after the White House issued its own AI Executive Order on June 1, and it lands in a regulatory environment where states have been filling the governance vacuum left by the absence of federal AI law. California, Colorado, Illinois, and New York have all enacted or are advancing their own AI rules. The Great American AI Act would impose a three-year pause on new state AI development regulations, betting that federal uniformity will prove more conducive to innovation than a patchwork of state-level requirements.

Feature Overview

The State Preemption Provision

The most contentious element of the bill is a three-year prohibition on new state laws regulating how AI models are developed — covering the training, building, and weighting of frontier models. States would retain authority to regulate AI at the deployment and use layer, meaning rules about algorithmic discrimination in hiring, housing, or credit could still be enacted. But laws governing the development process itself, including California's AB 2013 (requiring public training-data summaries) and portions of SB 942 (mandating AI content watermarking), would be preempted.

Sponsors frame this as a compromise that preserves consumer protection at the application layer while preventing fragmented training-side rules from driving frontier AI development offshore or into smaller, less regulated jurisdictions. Critics read the same provision as an effective elimination of the most substantive safety guardrails currently in place in the US.

Center for AI Standards and Innovation

The bill formally codifies the Center for AI Standards and Innovation within the Commerce Department and authorizes $100 million annually for fiscal years 2027 through 2029. The center would be responsible for developing voluntary guidelines, best practices, and security standards for AI systems, evaluating deployed systems, and monitoring AI development trends. Commerce would be required to publish its first voluntary AI governance framework within eighteen months of the bill's enactment.

Mandatory Audits and Incident Reporting

Large frontier AI developers — defined as companies exceeding $500 million in annual revenue — would face semi-annual third-party audits. Non-compliance carries penalties of up to one million dollars per day. Separately, covered companies must report critical safety incidents to the federal government. The bill does not define "critical safety incident" in the discussion draft, leaving that specification to rulemaking.

Government Accountability Requirements

The Government Accountability Office would be required to assess federal AI adoption progress, with particular attention to how agencies are complying with existing AI-related statutes. The Census Bureau and Bureau of Labor Statistics would need to incorporate AI use and adoption questions into federal surveys to build a baseline dataset on AI's workforce impact. The bill also directs identification of statutes and regulations that "unduly burden AI infrastructure," including energy infrastructure — a signal that the sponsors see permitting and power supply as downstream AI policy issues.

International Standards and China Exclusion

The Energy Department and NIST would jointly lead US engagement on international AI standards, with an explicit mandate to "form coalitions with like-minded governments" and an equally explicit exclusion of China from those coalitions. The provision aligns with the broader US strategy of building AI supply chain and standards agreements among democratic allies while treating Chinese standards bodies as adversarial participants.

Open-Source Security Grants

CISA would award grants to maintainers of widely used open-source software for security enhancements, including patching, maintenance, and audits. This provision is comparatively uncontroversial and reflects a bipartisan recognition that open-source AI infrastructure is a critical security surface.

Usability Analysis

For AI companies, particularly those headquartered in states with active AI legislation, the bill's passage would create meaningful legal clarity. A single federal compliance regime, even with semi-annual audits and incident reporting, is less operationally complex than maintaining separate compliance programs for California, Colorado, Illinois, and potentially a dozen additional states. The per-day penalty structure, however, raises the stakes of any compliance gap — a company that discovers a missed filing obligation late could face significant liability before remediation.

For state regulators and consumer advocates, the bill's preemption scope is the central problem. The three-year pause is not permanent, but it removes pressure from Congress to pass stronger protections — if industry can achieve three years of federal governance primacy with the promise of re-evaluation, the practical result may be a substantially weaker regulatory baseline than the states have been building toward.

For researchers and open-source contributors, the CISA grant program and the NAIRR codification are positive developments that increase public investment in AI infrastructure and research access.

Pros and Cons

Pros:

• Creates the first comprehensive federal AI governance framework in US history
• Eliminates state-level regulatory fragmentation that increases compliance costs for AI developers
• $100M annual Center for AI Standards and Innovation investment provides dedicated federal AI oversight capacity
• Semi-annual audits with real financial penalties ($1M/day) represent enforceable accountability
• Mandatory critical-incident reporting creates a federal early-warning mechanism
• NAIRR codification and CISA open-source grants increase public AI investment

Cons:

• Three-year state preemption eliminates California AB 2013 and SB 942, removing existing consumer protections
• AFL-CIO, Public Citizen, and House Democrats all oppose the draft in strong terms
• "Critical safety incident" remains undefined, deferring the most consequential definitional question to rulemaking
• Preemption applies to development-layer rules — exactly where frontier safety advocates believe regulation is most needed
• No definition of "frontier AI" in the draft, creating uncertainty about which companies are covered

Outlook

The Great American AI Act is a discussion draft, not introduced legislation, so its path to enactment is long and contested. The strong opposition from organized labor, consumer groups, and Democratic caucus members in the House signals that the preemption provision will require significant revision before the bill can attract the broader support needed for passage. The sponsors have structured it as a starting point for negotiation, soliciting written feedback rather than pressing for an immediate vote.

The competitive dynamic with state legislatures adds urgency on both sides. Colorado's comprehensive AI Act takes effect June 30, 2026 — twenty-three days from the draft's release date. If Congress cannot pass a federal framework before states accumulate a critical mass of AI law, preemption becomes a harder political argument to make, because it requires actively removing protections that are already in effect rather than simply establishing federal primacy over a blank slate.

The bill's most durable provisions are likely to be the ones with broad bipartisan support: NAIRR codification, CISA open-source grants, and the international standards mandate. The preemption fight will determine whether the broader framework survives or whether Congress ends up passing a narrower bill that leaves state law largely intact.

Conclusion

The Great American AI Act represents the most substantive federal AI governance proposal to reach public debate in the United States. Its bipartisan authorship and 269-page scope signal serious legislative intent, but the three-year state-law preemption provision is a fault line that separates industry supporters from labor, consumer, and state-rights opponents. The outcome of that fight will shape the regulatory environment for AI development in the US for years. Stakeholders across the spectrum have until formal introduction to influence the draft through the public comment process. Rating: 3/5.

Editor's Verdict

Great American AI Act: Bipartisan Bill Proposes Three-Year Freeze on State AI Laws is a workable proposition that fills a clear gap, even if it doesn't fundamentally change the landscape.

The strongest case for paying attention is first comprehensive federal AI governance framework in US history, providing national legal clarity, which raises the bar for what readers should now expect from peers in this space. Reinforcing that, eliminates multi-state compliance complexity for AI developers operating nationally adds practical value rather than just headline appeal. The broader signal worth registering is straightforward: the three-year state preemption is not merely procedural — it specifically eliminates California AB 2013 and SB 942, removing the two most substantive US AI transparency laws currently on the books. On the other side of the ledger, three-year state preemption eliminates California AB 2013 and SB 942 consumer protections already in law is a real constraint, not a marketing footnote, and it should factor into any serious decision. Layered on top of that, opposed by AFL-CIO (15M members), Public Citizen, and House Democratic caucus leadership in strong terms narrows the set of teams for whom this is an obvious yes.

For AI industry watchers, strategy teams, and decision-makers tracking platform shifts, the smart move is to track its trajectory and revisit once the rough edges are filed down. For everyone else, the safer posture is to monitor coverage and revisit once the use cases that matter to your team are demonstrated in the wild.