AI Talk
The 10 year federal moratorium on state AI legislation was finally removed from the One Big Beautiful Bill Act (to many lawmakers’ relief). In this post, I provide more background on how the moratorium works, attempts to revise it, and how companies should approach state regulations in AI.
The AI moratorium was a part of President Trump’s domestic policy agenda that was included in the budget reconciliation measure called the One Big Beautiful Bill Act. The House passed the Act in May of this year, which included a broad 10-year federal moratorium on state regulation of AI systems, AI models, and automated decision systems. The rationale was to eventually create a unified, federal level approach to AI development and deployment by first preventing states from enacting their own AI laws, which, some argued, created a patchwork of regulations for businesses to navigate. The objective for this moratorium was to encourage innovation and reduce compliance burdens for businesses operating across state lines. Tucked away in the Act, it flew under the radar for a while until lawmakers started reviewing it more closely.
The broad moratorium drew bipartisan controversy across the political spectrum, with senators particularly concerned with the Act potentially undermining protections in children’s online safety, deceptive trade practices, and civil rights, with the proposed revisions not adequately addressing these areas while introducing new ambiguity. And, given the high speed of technology development and evolving use cases, a ten-year moratorium was too long in duration.
In an effort to pass the moratorium, it was revised a few times. First, the duration was shortened from 10 years to 5, and clarified that only the newly earmarked $500 million in federal funding would be contingent on state compliance with the moratorium. Then, the “generally applicable law” exception was revised to still allow state regulation over unfair or deceptive acts or practices, online child safety, and protections against non-consensual use of ones’ name, image and likeness, reflecting concerns from certain senators. However, this revised exception created new legal uncertainty and ambiguity by requiring that the state laws also not impose an “undue or disproportionate burden” on AI systems, and that they “reasonably effectuate the broader underlying purposes” of the legislation.
The final vote, 99-1 on July 1, 2025, decisively removed the moratorium from the Act, but it illustrates the challenges of creating a cohesive national AI policy, while allowing states to regulate their own businesses and protect their citizens, and while still encouraging innovation in AI technology. In the meantime, businesses should prepare to implement and comply with existing and new state laws coming into effect. In prior posts, I covered AI legislation in California (SB 942 and AB 2013) and Colorado, and in my next post, I discuss the Texas Responsible AI Governance Act (TRAIGA), which will come into effect on January 1, 2026. Stay tuned!
