Regulation and Policy

AI regulation in 2026 is a patchwork stitched from four different philosophies. The European Union passed the world's first comprehensive AI statute, the AI Act (in force 1 August 2024), which bans a short list of practices, sets duties for "high-risk" systems and for general-purpose models, and fines up to 7% of global turnover; in July 2026 it then delayed the high-risk rules to December 2027 and August 2028 through a "Digital Omnibus" amendment. The United States has no federal AI statute. Its federal government moved from the safety-oriented executive order of October 2023 to a deregulatory, "win the race" posture in 2025, and in December 2025 ordered federal agencies to challenge state AI laws, while the states themselves passed hundreds of AI bills, including California's frontier-model transparency law (SB 53, in force January 2026), New York's RAISE Act (effective January 2027), Texas's TRAIGA (January 2026) and Colorado's pioneering, repeatedly delayed and finally rewritten AI Act. China regulates AI through a fast-moving sequence of targeted rules administered by the Cyberspace Administration of China, covering recommendation algorithms, deepfakes, generative AI, content labelling and, from July 2026, human-like "companion" AI, with security and content control at their core. The United Kingdom chose a "pro-innovation" approach that relies on existing regulators and a well-funded AI Security Institute, and had still not introduced a promised frontier-AI bill by late 2026. This section explains each regime, the fights over federal preemption in the US, the economic logic of AI regulation, what compliance actually costs and involves, and how an experienced policy or compliance lead reads the next move.