The Three-Jurisdiction AI Compliance Matrix — EU Enforcement Now, UK Principles-Based, US Patchwork State-by-State

The August 2 EU AI Act enforcement date crystallises the divergence between three regulatory models that transatlantic law firms are now navigating simultaneously. The EU model is rules-based and prescriptive: Article 50 transparency obligations are in force with a live penalty framework (up to EUR 15 million or 3% of worldwide turnover), the AI Office has enforcement powers against GPAI providers, and the Annex III high-risk timeline runs to December 2027. The UK model remains principles-based within the existing regulatory framework: the AI Growth Lab for legal services provides a voluntary sandbox, the BSB's May 2026 guidance recontextualises existing professional obligations, and the new automated decision-making code of practice (SI 2026/425) operates within the Data Protection Act structure. The US model is a patchwork of state-level statutes — Connecticut's CART Act (October 2026), Illinois' AI interview law, New York's automated employment decision rules, Colorado's AI Act — with no federal framework in force, and bar association ethics opinions (Alabama, New York, California) establishing professional conduct obligations in the absence of legislation. For general counsel at multinational organisations and for law firms with transatlantic practices, the practical consequence is a three-track compliance architecture: EU-facing AI tools require Article 50 disclosure infrastructure from August 2; UK-facing tools require alignment with automated decision-making documentation and BSB/SRA guidance; US-facing tools require state-by-state mapping and bar ethics compliance. The jurisdictions are not harmonising; the divergence is structural and is widening with each quarter's new state statute and EU implementing measure.