What happened
A Center for the Governance of AI study, analysing 375 large language models released between June 2018 and May 2026, found 11% of model releases from companies including Meta, Google, OpenAI, and Anthropic were delayed or unreleased in the EU compared to the US; 7% were similarly affected in the UK. Of 68 identified delays and non-releases, regulatory factors caused 56 cases. Data protection regulations are the primary barrier, particularly for non-text modalities. Meta experienced the highest rate, with 26% of its releases delayed or withheld in the EU and 15% in the UK, while Claude 3 Opus web app saw a 71-day EU delay.
Why it matters
European businesses face delayed access to frontier AI models, impacting competitive parity and innovation adoption. Data protection regulations, particularly GDPR, and the EU's aggressive enforcement create a constraint on model deployment, especially for non-text modalities. This mechanism limits advanced AI tool availability for platform engineers and founders, a pattern seen as AI adoption outpaces governance globally. Procurement teams must factor in regional regulatory friction when planning AI infrastructure, anticipating potential delays and reduced model choice.
