Download this commentary (PDF)
In the absence of a federal approach to AI regulation, U.S. states have introduced nearly 1,200 bills and enacted 150 into law.
The landscape of US regulation on AI is more complex than is generally acknowledged.
We saw that recently, when, on July 6, 2026, Governor JB Pritzker of Illinois signed into law the Artificial Intelligence Safety Measures Act (SB 315). The Illinois law requires AI labs to have their safety practices audited by an independent third party — a legal requirement no other jurisdiction in the United States imposes. Yet Illinois did not act in isolation. Its law is the third element of what appears to be an intentionally coordinated set: California's SB 53 covers frontier-developer transparency, New York's Responsible AI Safety and Education (RAISE) Act combines incident-reporting with an oversight office, and the new statute adds the audit layer. Using shared scope, definitions and core obligations, the cumulative effect of the three States' laws is to create a coherent compliance regime — and, surprisingly, to pull U.S. AI regulation toward a legal construct comparable to the EU's AI Act.
Federal Regulation vs. Regulation by the States
The Trump administration has sought to deregulate at national level and, at the same time, to preempt the States from regulating AI. Federal AI policy has focused on revoking state-level AI policies and directives that act as barriers to innovation, codified in Executive Order 14179 of January 23, 2025, “Removing Barriers to American Leadership in Artificial Intelligence.” Preemption of state law on AI originally had been part of the One Big Beautiful Bill Act (OBBBA), a bundle of second term priorities for the Trump administration, but the Senate voted 99–1 to remove it in July of 2025.
In the absence of hard law at the federal level, state legislatures introduced over 1,200 AI-related bills in 2025 and enacted almost 150 into law. The evident failure of the moratorium prompted the President to adopt the “one rule” EO 14365 of December 11, 2025, which uses a ‘stick’ approach—executive-branch levers which could lead to withholding federal funding—to prevent States legislating “in conflict with national policy priorities.” The measure succeeded in rolling back or narrowing legislation in several States, including Colorado, Utah, and Texas, but it has not stopped the complementary legislation passed in California, New York and Illinois.
Converging Policy Trends on Both Sides of the Atlantic?
Meanwhile, in Congress, a bipartisan duo are proposing to take the New York-Illinois-California regime on AI and make it federal. This is the essence of the discussion draft floated on June 4, 2026, by Representatives Jay Obernolte (R-CA) and Lori Trahan (D-MA), titled the “Great American AI Act” (GAAIA).
Yet the GAAIA does more than synthesizing state laws. It would also establish a comprehensive product safety compliance regime. This element of the framework is classic U.S. product-safety legislation, in the tradition of the Consumer Product Safety Act. It is a consumer-protection law that attempts to correct market failures rather than proclaiming rights. To this end, the GAAIA requires transparency about risk mitigation and safety incidents to the Director of the Center for Artificial Intelligence Standards and Innovation (CAISI), which is the federal body that licenses independent verification organizations to audit large model development.
These developments point in a similar direction to the EU. While the EU AI Act is generally considered a “rights-based” or “precautionary” instrument — grounded in the protection of rights such as privacy, non-discrimination and human dignity — in contrast to a “market-based” American approach. In fact, the EU AI Act follows a standard template which is familiar to product-safety lawyers.
The similarities between the two are striking. Each begins by defining what is covered and who is responsible: the GAAIA defines the class of frontier models and the developers subject to the regime, while the EU AI Act identifies a defined the high-risk AI system (“the product”) and a chain of economic operators (provider, importer, distributor, deployer). Each requires the AI system to be checked against safety requirements before it reaches the market. Under the GAAIA, this step involves a third-party assessment; in the EU, on the other hand, a manufacturer may self-declare that its product conforms with the regime’s requirements. Each mandates the reporting of adverse events or serious incidents to a public authority, backed by enforcement.
With respect to compliance, it bears emphasis that the US requirement is the more onerous of the two. Whereas the GAAIA requires obligatory assessment by an independent third party, the EU regime relies on the manufacturer’s own declaration of conformity.
The fit is not totally perfect. The EU AI Act has a special carve-out for general purpose AI, which is more of a service and does not fit the product mold. Moreover, the EU AI Act arguably elides serious discussion of fundamental rights. Finally, the narrower approach on the US side may be more sustainable in the longer term. Nevertheless, in broad terms, the comparison holds.
A De Facto National Standard?
Despite resistance from the federal level, the three states California, New York and Illinois may well shape US policy going forward, for two reasons. Not only are they significant economic powerhouses in their own right, but many of the companies building frontier models are likely to comply with the strictest common standard wherever their customers are located — much as global companies chose to comply with the EU's General Data Protection Regulation (GDPR) worldwide rather than run separate systems for separate markets. Secondly, Congress is taking notice: the Trahan-Obernolte draft would federalize the states' approach, converting their coordinated template into a single framework, and preempt other state law.
Should federal legislation begin to crystallize around a template pioneered by the states — one that Brussels would recognize — it is not inconceivable that we may yet witness a United States whose regulatory reach extends as far as its economic power. Illinois has just given us a hint of how.
