AI governance is fragmenting
Experts say regional AI rules are splintering and that governance is becoming infrastructure — a shift that’s pushing enterprise AI from experiments into operations and complicating cross‑border deployments. (x.com)
The EU’s Artificial Intelligence Act entered into force on August 1, 2024 and uses a phased timetable that makes most high‑risk obligations applicable on August 2, 2026. (commission.europa.eu) China’s “Interim Measures for the Management of Generative AI Services” were issued jointly by the Cyberspace Administration and six ministries and took effect on August 15, 2023, requiring providers to label AI‑generated content and keep filings and records with authorities. (regulations.ai) The U.S. has relied on executive action and agency guidance—President Biden’s October 30, 2023 Executive Order (E.O. 14110) directed more than 50 federal entities and more than 100 follow‑on actions—while later White House directives in December 2025 sought a national AI policy framework. (bidenwhitehouse.archives.gov) Surveys show enterprises are shifting AI from pilots into production: McKinsey found roughly three‑quarters of firms use AI in at least one function and only about one‑third say they are scaling across the enterprise, while Gartner reported 45% of high‑maturity organizations keep AI projects operational for three years or more. (kanerika.com) Industry analysts and vendors report governance tools are being embedded into engineering stacks—Cisco’s January 2026 study says organizations are investing in data governance as core infrastructure, and the World Economic Forum has argued governance is now “infrastructure” for physical AI systems. (newsroom.cisco.com) Consultancies and vendors document concrete integration work: MLOps guides and vendor surveys show firms are wiring compliance checks into model pipelines, and one industry note cited data‑quality losses at about US$12.9 million per year and predictive downtime at roughly US$125,000 per hour as drivers for that integration. (elevateconsult.com) The result is rising cross‑border friction: the EU Act explicitly places obligations on third‑country providers whose outputs are used in the EU, China’s measures include foreign‑investment and filing rules, and recent surveys flag security, integration and regulatory compliance as top barriers to multinational AI rollouts. (artificialintelligenceact.eu) Regulatory timelines and penalties are tightening the business case for built‑in governance—EU non‑compliance can trigger fines up to €35 million or 7% of global turnover, and phased enforcement milestones through 2026–2027 mean multinational vendors must map differing obligations now. (gdprlocal.com)
