EU opens first AI Act enforcement cases against OpenAI and Anthropic

By Splitfeed EditorialLast updated Oct 15, 2026, 08:30 UTC

On October 15, 2026, the European Union opened its first formal enforcement cases under the AI Act, naming OpenAI and Anthropic as the initial targets. The cases mark the first time the regulation's enforcement mechanism has been used since the law came into force. No specific allegations or demands have been detailed in the available sources, but the fact that both of the most prominent large language model developers are simultaneously in the crosshairs signals that regulators chose to open with high-profile targets rather than smaller or lower-risk actors.

The AI Act was designed as a risk-tiered framework, with the heaviest obligations falling on systems deemed high-risk or, at the frontier, general-purpose AI models above certain capability thresholds. OpenAI and Anthropic each develop models that fall under the most scrutinised category. Enforcement cases do not automatically mean fines or sanctions, but they do trigger formal investigation procedures that can compel companies to hand over documentation, model evaluations and internal assessments. What remains genuinely in dispute is whether the Act's requirements, written before the current generation of frontier models matured, are technically precise enough to produce consistent rulings, and whether EU enforcers have the technical capacity to evaluate compliance meaningfully.

None of the three camps had published reactions at the time of writing. Pro-AI voices would typically argue that enforcement actions of this kind risk pushing frontier AI development outside Europe and burdening compliant companies with obligations that slower-moving competitors elsewhere do not face. Anti-AI voices would typically welcome the cases as long-overdue accountability, arguing that companies building potentially high-impact systems should face structured scrutiny before harms materialise. Middle-ground observers would typically call for precision, urging enforcers to focus on specific, demonstrable risks rather than broad investigations that could produce uncertainty across the industry without improving safety outcomes.

The details worth watching are the formal scope of each investigation when published, any interim compliance demands the EU issues, and whether other frontier developers receive similar notices in the weeks that follow. A first ruling or settlement in either case would set a precedent for how the AI Act functions in practice, and its terms would clarify which obligations regulators consider most urgent.