EU tech chief says the bloc is well equipped for very capable AI models

Henna Virkkunen, the European Commission’s Executive Vice-President for tech sovereignty, security and democracy, told Reuters on Friday that EU rules can deal with risks from very capable models. In an interview carrying a Brussels dateline of 9 October, syndicated by Global Banking and Finance Review, she said: “We see that the safety and security of very capable models is a very hot topic internationally and we in Europe are well equipped for that.” She added: “We have our AI Act in place and the AI Act covers the whole life cycle of these models.”

What the Commission asked for

The Reuters copy says the Commission sent requests for information in late August to more than 30 AI companies, asking about safety and security measures, then followed up on transparency and copyright. Reuters says the Commission did not name the companies, and that Virkkunen is now assessing the responses. Asked about Chinese AI startups, she said the Commission had “also sent requests for information to them,” and that the “most capable models” now on the market were from the United States and China.

MLex’s headline and visible standfirst, by Masha Borak and dated 9 October 2026, say Virkkunen told MLex that OpenAI, Anthropic, and Meta Platforms are among more than 30 companies asked about compliance, including safety and copyright. The rest of that article is behind a subscription wall. The names are MLex’s account of what she told them. Reuters says the Commission did not name recipients.

What general-purpose model duties require

The duties are on the Commission’s site, separate from Friday’s interview. Guidelines for providers of general-purpose AI models say those obligations entered into application on 2 August 2025, and that the Commission’s enforcement powers enter into application on 2 August 2026, including fines. Providers of models placed on the market before 2 August 2025 must comply by 2 August 2027.

A Commission FAQ lists the duties for providers: technical documentation for the AI Office; information so downstream providers can understand capabilities and limits; a policy to comply with Union copyright law; a sufficiently detailed public summary of training content; and, for a provider established outside the EU, an authorised representative in the Union before the model is placed on the market. Models the Act treats as carrying systemic risk have further duties, including notification, evaluations, incident reporting, and cybersecurity. The same FAQ says that from 2 August 2026 the Commission enforces full compliance, including through fines.

The enforcement page says the AI Office can send requests for information to check compliance. For a simple request, it says, fines can follow an incorrect or misleading reply. For a request issued by Commission decision, fines can also follow a failure to reply or an incomplete reply. Prohibited-practice breaches, it says, can reach 35 million euros or 7 percent of worldwide annual turnover, whichever is higher. Other breaches, including the obligations for general-purpose models, can reach 15 million euros or 3 percent.

Reuters says these information requests can lead to investigations and fines of as much as 7 percent of global annual turnover. The wire does not mention the 3 percent ceiling the Commission page states for general-purpose-model breaches. The 7 percent figure in the wire matches the prohibited-practice maximum on that page.

OpenAI has already described, in a 5 October post covered here, an approach to EU rules on marking generated text. Nothing in Friday’s interview says the information requests are about that programme.

Virkkunen also told Reuters she rejects the idea that the rules are already outdated, pointing to risk assessment, outside experts, and continued monitoring of models. She said regulators are guiding companies on how to evaluate models and how much time to allow, and that recommendations from a scientific panel of 60 AI experts would be factored in.