AI rules are fragmenting
National approaches to AI governance are diverging fast, creating a patchwork that makes cross-border enforcement and oversight of general-purpose models increasingly fraught — regulators’ priorities and legal traditions simply don’t line up. Analysts warn this fragmentation will raise compliance costs and operational complexity for firms that deploy AI globally, while the Global South pushes alternative sovereignty models to protect local data and capability. (theregreview.org) (moderndiplomacy.eu)
The EU’s AI Act phases obligations across three dates: rules for general‑purpose AI began applying on 2 August 2025, the bulk of requirements and national enforcement mechanisms come into force on 2 August 2026, and the full roll‑out is scheduled by 2 August 2027. (ai-act-service-desk.ec.europa.eu) The U.S. now shelters under a December 11, 2025 executive order that seeks a single federal AI framework while dozens of state AI statutes — including California’s and others that took effect on January 1, 2026 — create a legal clash likely to be resolved in courts. (whitehouse.gov) China has embedded AIGC and “deep synthesis” controls into national rules that require watermarking, labelling and sectoral oversight for AI content providers and has layered these on top of cybersecurity and data rules. (practiceguides.chambers.com) Analysts quantify the price of fragmentation: Gartner projects AI‑governance spending at $492 million in 2026 and estimates the market will exceed $1 billion by 2030, while regional estimates put initial EU AI Act compliance for some small firms at roughly €320,000–€600,000. (gartner.com) Multinationals warn that the EU’s extraterritorial reach will force non‑EU providers to build separate governance controls for EU residents, even as the U.S. executive order directs the Commerce Department to identify and challenge “burdensome” state laws by March 11, 2026. (ai-act-service-desk.ec.europa.eu) The Global South is formalizing alternative “sovereign AI” models: India’s AI Impact Summit (16–20 Feb 2026) produced a declaration signed by 89 signatories and a set of governance guidelines that report India has provisioned 38,000 GPUs, hosts 9,500 datasets in AIKosh, and runs 40+ petaflop national supercomputing capacity. (government.economictimes.indiatimes.com) Regional blocs are following suit — the African Union adopted a Continental AI Strategy in July 2024 and is moving to harmonize data‑sovereignty rules, while Brazil has launched a national AI plan stressing digital sovereignty and a multibillion‑dollar investment push. (au.int)
