What happened
The Centre for the Governance of AI (GovAI) published a research paper analysing 375 LLM releases by Meta, Google, OpenAI, and Anthropic from June 2018 to May 2026. Key findings: 11% of model releases were delayed or not released to the EU; 7% delayed or not released to the UK. Meta had the highest rate (~26% EU delays). Regulatory barriers — primarily GDPR data-protection requirements, not the EU AI Act — are identified as the primary cause of most delays. The study establishes an empirical baseline ahead of the EU AI Act's GPAI (general-purpose AI) provisions entering force in August 2026.
Why it matters
The paper is directly policy-relevant: it documents that the EU's data-protection framework is already materially restricting frontier AI market access — and signals the AI Act's August 2026 GPAI obligations will add a second regulatory layer on top. For AI deployers and the European Commission, it frames the compliance cost of data-protection rules and sets expectations for further delays when GPAI rules take effect.
Action needed
EU-market AI deployers should audit GDPR compliance gaps that could trigger delayed access to new models post-August 2026 GPAI rules. Track European Commission guidance on GPAI classification (draft published May 19, 2026) as the consultation closes.