GDPR Cited as Key Barrier to AI Releases in Europe

A recent Center for the Governance of AI report indicates that leading AI companies have delayed or withheld some large language model releases in the European Union more often than in the United States. The research reviewed 375 model releases between June 2018 and May 2026 by companies including Meta, Google, OpenAI, and Anthropic.

According to the report, at least 11% of relevant releases were delayed or unavailable in the EU compared with the United States, while 7% faced delays or non-release in the United Kingdom. Of 68 identified cases, the researchers attributed 56 primarily to regulatory factors. Claude 3 Opus, for example, reportedly reached EU users through its web application 71 days after its US release.

The General Data Protection Regulation appears to be the regulatory framework most frequently associated with these decisions. The report suggests that providers face particular difficulty when launching AI products involving images, audio, and live video, rather than text-only functions. It attributes the higher apparent EU impact, compared with the UK, to more active enforcement and continuing uncertainty over the application of data protection rules to AI training and deployment.

The Digital Markets Act and AI Act may also affect future market access, although their practical effect remains less clear because their relevant obligations have only recently begun to apply. At the same time, EU institutions are considering changes intended to make certain data rules more workable for AI development, while reviewing copyright rules and the AI Act’s copyright-related provisions. The policy challenge is to provide legal certainty and protect fundamental rights without creating unnecessary barriers to the availability of lawful AI services in the EU.