Regulation

On 1 July 2025 the US Senate voted 99–1 to strip from the budget bill a provision that would have imposed a ten-year moratorium on state laws regulating artificial intelligence (Senate record). Federal preemption was the position of the administration, of Senator Ted Cruz, of much of the technology industry and of some venture investors, including Andreessen Horowitz, who argued that a "patchwork" of fifty state regimes would raise compliance costs and risk ceding the race to China. Against them stood state legislators and attorneys general of both parties, governors including Gavin Newsom, and consumer and child-safety groups, who argued that preemption without a federal standard was simply deregulation. The vote showed that preemption without a federal standard lacked support even in the majority party.

The vote did not settle the question. A month later, on 2 August 2025, the European Union's obligations for general-purpose AI models began to apply under the AI Act, the world's first comprehensive AI statute. On 29 September 2025 California's governor signed SB 53, the first US law aimed specifically at frontier-model developers. On 11 December 2025 President Trump signed Executive Order 14365, which directed the Justice Department to challenge state AI laws in court. By late 2026 an AI company selling worldwide faced a binding EU statute that had delayed its own hardest rules, a growing body of US state law under federal attack, and a separate, faster Chinese regime built around content control.

This chapter explains the main regimes, what they require, what compliance costs companies of different sizes, and why the rules matter to the economics of AI even when they change.