The Great American AI Act Trades a Patchwork for a Federal Ceiling

The Core Trade

On June 4, 2026, Rep. Jay Obernolte (R-CA) and Rep. Lori Trahan (D-MA) released a discussion draft of the Great American AI Act, bipartisan legislation to create a federal framework for how the United States governs artificial intelligence. The draft runs 269 pages. The operative sentence for operators is much shorter: "No state or political subdivision thereof may establish, continue in effect, or enforce any law or regulation specifically regulating the development of any artificial intelligence model."

That clause freezes state-level AI development regulation for three years—the sharpest preemption language yet to emerge from a bipartisan House coalition. It arrives after a year of failed attempts through budget reconciliation and defense authorization bills. The pressing question is whether the federal floor being offered justifies that state ceiling.

What the Bill Actually Does

The discussion draft creates four pillars for AI advancement: establishing frontier AI model governance, collecting insight into changes within the U.S. workforce landscape, fortifying cybersecurity postures, and spurring new AI research and development.

The preemption mechanics matter most: the bill would preempt state laws and regulations "specifically regulating the development" of an AI model, with a three-year sunset, and the bill specifies that preemption would not apply to laws related to the use or deployment of AI models. This development-versus-deployment distinction is the fault line for state attorneys general, consumer advocates, and enterprise compliance teams. States keep authority over post-deployment. They lose authority over training.

On developers themselves: the bill would require large frontier developers — those with more than $500 million in gross revenue for the previous calendar year — to establish public frontier AI frameworks. These frameworks must showcase standards compliance efforts, identify risk thresholds, determine whether a model poses "a catastrophic risk" when managing cybersecurity defenses, and disclose release dates.

Developers would report incidents within 15 days of discovery and imminent risks within 24 hours. The bill also requires transparency in frontier AI, independent verification organization audits and assessments, and anti-retaliation protection for AI whistleblowers.

For institutional infrastructure: the bill would create the Center for AI Standards and Innovation (CAISI), an office in the Commerce Department's National Institute of Standards and Technology, to evaluate frontier models for the next three years. CAISI would receive $100 million annually for 2027 to 2029.

The State Law It Displaces

The bill targets specific casualties. An accompanying document released by Trahan's office said that California's AB 2013 law, which "requires model developers to publicly post high-level summaries of their training data," would be preempted, along with a portion of California SB 942 related to content watermarking. It also called out frontier safety laws in California, New York and Illinois, saying they would be "federalized" under the bill.

California SB 53 is most directly in the bill's crosshairs. By passing SB 53, California became the first state in the U.S. to directly regulate developers of "frontier" foundation models — that is, companies that develop AI models trained with more than 10^26 floating-point operations. Under that law, large frontier developers must publish and keep current an enterprise-wide "frontier AI framework," release detailed public transparency reports summarizing model capabilities and risk-assessment results, submit rolling summaries of internal catastrophic-risk assessments, report any "critical safety incidents" to the California Office of Emergency Services within specified time periods, and face civil penalties of up to $1 million for violations enforced by the California Attorney General.

The Great American AI Act mirrors SB 53's structure—public frontier AI frameworks, incident reporting, whistleblower protections—but federalizes enforcement, removes the California AG's direct authority, and relegates state attorneys general to an opt-in reporting channel. State attorneys general could opt in to receive safety reports. This is a demotion.

Colorado's AI Act, which takes effect on June 30, 2026, imposes various obligations related to documentation, disclosures, and governance of "high-risk" AI systems — systems that make "consequential decisions" relating to education, employment, healthcare, and similar areas. The bill's preemption of development-side regulation creates ambiguity for Colorado since its law covers both developers and deployers.

The Fragmentation Problem the Bill Solves

The preemption argument rests on solid ground. Across 2025, more than 70 AI-related laws passed in at least 27 states, illustrating the rapid evolution of state-level AI governance. According to the National Conference of State Legislatures, all 50 states and the District of Columbia introduced some form of AI legislation in the 2025 legislative session. For a frontier lab or enterprise platform provider operating nationally, that means 50 parallel compliance tracks with divergent definitions, thresholds, enforcement mechanisms, and timelines.

The bill's sponsors frame the stakes directly. "Rather than allow protections to exist in only a handful of states or force innovators to navigate dozens of different legal regimes, our framework would establish one national standard," Obernolte and Trahan wrote in an op-ed in Bloomberg Law.

The Critics' Strongest Argument

Opposition extends beyond advocacy groups. Brad Carson, president of Americans for Responsible Innovation and a former Democratic representative from Oklahoma, called preemption a "generational mistake." "This bill takes the current floor on state AI legislation and turns it into a federal ceiling, preventing state lawmakers from addressing emerging AI harms in an era of fast-moving technology," he said.

According to the Tech Oversight Project, 56% of Obernolte's constituents oppose his "efforts to weaken California's AI laws governing catastrophic risk," and 63% of Trahan's constituents feel the same way about local laws.

Their strongest point is definitional. This bill's compliance burden depends entirely on how "frontier model" is defined in final text and whether the risk-management framework requirements have teeth or amount to disclosure theater. While federal preemption eliminates tracking multiple state regimes, it imposes its own substantial compliance burden. Organizations should not assume that federal preemption translates to reduced compliance expenditure; it will more likely merely signal a shift in where those resources are directed.

This remains a discussion draft. It is intended to solicit feedback from stakeholders, experts, and the public before formal introduction. Definitions will shift. Thresholds will be lobbied. The $500 million revenue cutoff for "large frontier developers" already mirrors SB 53's trigger—that is no accident, and it will not survive markup unchanged.

What to Watch

1. Frontier model definition. The bill's scope hinges on the compute threshold. Whether it matches SB 53's 10^26 FLOPs or moves lower determines whether mid-tier labs fall under the mandate.

2. Risk-management specificity. Vague framework requirements produce theater. Watch whether CAISI rulemaking sets prescriptive process requirements or simply asks for a document.

3. The development-versus-deployment line. The draft preempts states from issuing their own laws regulating frontier AI model development, but allows states to pass laws of "general applicability" related to AI and regulate models after deployment. This boundary will be contested in markup and court.

4. Open-source carve-outs. The bill text does not yet specify whether open-weight model releases fall under the frontier developer mandate. If they do, the compliance cost lands on labs that release weights publicly—structurally different from closed-API providers.

5. Senate arithmetic. If revived outside reconciliation, preemption efforts would likely require 60 votes in the Senate, necessitating bipartisan cooperation. The House coalition is real. The Senate path is not.

6. State AG response. California's attorney general retains enforcement authority under SB 53 until preemption takes legal effect. Watch for a constitutional challenge before the bill reaches a Senate floor vote.