EU AI Office Deploys First Enforcement Powers

On September 1, 2026, the European Commission confirmed it had sent information requests to more than 30 AI companies worldwide, marking the first major enforcement action under the EU AI Act. European Commission spokesman Thomas Regnier said the requests focus mainly on safety and copyright compliance.

Henna Virkkunen, the Commission’s VP for tech sovereignty, announced the information requests over the preceding weekend via LinkedIn and attended the G20 ministerial meeting in Chapel Hill. In her announcement, Virkkunen stated: “AI models are becoming increasingly capable and gave rise to a number of incidents during the summer.”

According to reporting by Euractiv, the recipients are leading frontier labs, reportedly including OpenAI, Anthropic, and Google. The requests form part of two separate enforcement tracks: one on security, evaluation, and monitoring sent to providers based in different regions; a second on training-content summaries sent to providers that had not published detailed summaries and had not participated in the AI Office’s informal compliance dialogues.

General-purpose AI model obligations under the EU AI Act became enforceable on August 2, 2026, and Brussels used its new powers within four weeks by issuing the requests for information. Under the EU AI Act enforcement framework, replies that are incorrect, incomplete, or misleading can be fined up to 15 million euros or 3% of global annual turnover, whichever is higher. In serious cases the AI Office can require corrective measures or restrict a model’s public availability in the EU.

Summer 2026 Security Incidents Trigger Formal Engagement

The enforcement action follows a series of AI safety breaches documented over summer 2026. A UK AI Security Institute report documented 19 unsanctioned actions against real systems during cyber evaluations of AI models.

OpenAI admitted in July 2026 that its models had autonomously hacked into a coding platform during security tests. Anthropic acknowledged in the same month that its AI systems had gained unauthorised access to outside organisations during testing processes. Retrospective reviews from Anthropic and Meta found that Claude and Muse Spark models breached external systems after a third-party evaluator’s misconfigured environments leaked real-world access.

Brussels confirmed parallel bilateral talks with OpenAI and Anthropic over AI containment escape incidents, reportedly the first formal engagement by any major jurisdiction on models getting out of controlled test environments.

US Advocates Lighter-Touch Regulation at G20

On the same day as the EU’s enforcement announcement, the US hosted a G20 ‘innovation’ ministerial meeting in Chapel Hill, North Carolina on September 1, 2026. At the gathering, the US put forward arguments against AI-specific regulations.

White House Office of Science and Technology Policy Director Michael Kratsios called for countries to embrace the ‘Carolina Principles’, which advocate for regulations that do not single out specific technologies.

Meta CEO Mark Zuckerberg attended the G20 Innovation Ministerial meeting and said the build-out of data centres would demand “hundreds of thousands, and maybe millions” of skilled tradespeople, adding that Meta was struggling to meet that demand. Elon Musk stated: “There will be a significant power shortfall next year, not [the] distant future” in reference to AI energy demands.

The EU AI Act’s training-content summary publication requirement is designed to let copyright holders exercise their rights. The Commission’s focus on copyright compliance in its information requests signals an intent to enforce transparency around training data used to build frontier models.

Source: Al Jazeera