The U.S. and EU are taking opposite stances on AI regulations.
A bipartisan group of federal lawmakers in June proposed largely freezing states’ abilities to regulate artificial intelligence (AI).Three days earlier, the European Union began enforcing a landmark piece of AI legislation designed to hold the world’s most powerful AI companies accountable. The contrast could not be sharper.
A discussion draft circulated in the U.S. House of Representatives, the Great American Artificial Intelligence Act of 2026 (GAAIA) arrived just days after President Donald J. Trump signed an executive order establishing voluntary federal agency reviews of new frontier AI models. Together, these efforts to regulate AI underscore that federal AI governance has become both a congressional and executive priority.
At its core, however, the debate is not simply about whether the United States needs a national AI framework. It is about what kind of framework it should build: a federal floor that preserves state innovation or a federal ceiling that displaces it.
The bill has two structural components that deserve separate analysis. The first is a federal governance framework targeting frontier AI developers above $500 million in annual revenue. These developers would be required to publish safety frameworks, report critical incidents, and submit to semi-annual third-party audits.
The bill would also formally establish the Center for AI Standards and Innovation within the National Institute of Standards and Technology, building on the entity that previously operated administratively as the AI Safety Institute. The new center would develop guidelines and evaluating both domestic and foreign AI systems.
On its own, this framework is a serious attempt at federal AI governance. The audit requirement, in particular, is more demanding than most state-level frameworks have imposed. For this reason, the Act could reasonably be described as progress.
The second component is harder to defend. The bill would preempt state laws specifically regulating the development of AI models for three years, while allowing states to retain authority over deployment and use.
That distinction sounds technical. In practice, it is the difference between regulating the conditions under which a weapon is built and regulating only where it can be fired.
The preemption clause arrives at a moment when state-level AI governance has become the most substantive regulatory activity in the United States. California has enacted multiple AI laws covering training data transparency, content detection, healthcare AI disclosures, and companion chatbot safety. Colorado revised its landmark AI law. New York enacted its own AI framework. These laws represent years of legislative deliberation and hard-won political consensus.
The federal bill’s supporters argue that preemption is necessary to avoid a confusing patchwork of state rules. This argument has surface plausibility; a company operating nationally should not need 50 compliance programs. But it conflates regulatory complexity with regulatory absence. The solution to a patchwork is a federal floor that preserves state innovation, not a federal ceiling that removes it.
Viewed as a whole, GAAIA presents the most comprehensive federal framework for frontier AI governance proposed to date. Yet its sweeping preemption provision ultimately undermines that achievement by suspending the very state-level experimentation that has driven AI governance in the United States.
The preemption clause converts a period of active democratic experimentation into a three-year regulatory void at the developmental layer—one which, given the bipartisan opposition the draft has already drawn, the U.S. Congress may ultimately narrow before any bill is enacted.
The timing of this bill cannot be analyzed in isolation. Three days before the bill’s release, the European Commission formally appointed a 60-member scientific panel and a 174-member advisory forum to support enforcement of the EU AI Act. The next major milestone in the EU AI Act’s phased implementation is already here.
The contrast is striking. The EU is empowering independent technical experts with the ability to scrutinize whether major AI providers’ compliance claims hold up. The sponsors of GAAIA would create a lighter federal framework and dismantle the state-level architecture that had been filling the governance vacuum.
For companies operating across both jurisdictions, this creates an immediate strategic question. The EU AI Act applies to any provider placing AI systems on the EU market, regardless of where the provider is based. A US company that builds to the minimum federal standard under GAAIA, while state-level protections are suspended, will still face the full weight of EU enforcement in Europe.
The question for AI companies is whether to adopt a separate, lighter compliance posture for the U.S. market during the preemption window or to harmonize globally with the EU’s higher standard. The companies that choose the former are making a significant set of assumptions about a bill that may not pass as at all.
A federal AI framework is both necessary and overdue. The patchwork argument is not wrong. It is simply being used to justify the wrong remedy. The right answer is a federal floor that sets minimum national standards while preserving state authority to go further, not a federal ceiling that suspends state innovation while federal standards are still being written.
The EU did not build its enforcement infrastructure by asking member states to stand down while the Commission figured out the rules. The United States can do the same without treating state-level democratic accountability as an obstacle to be preempted rather than a resource to be integrated.
GAAIA, in its current form, gets the first part right and the second part dangerously wrong. Whether the sponsors use the discussion-draft stage to narrow the preemption clause before formally introducing the bill will tell us a great deal about whose interests the final legislation is designed to serve.
