The European Commission has sent the first Artificial Intelligence Act requests for information, or RFIs, to more than 30 AI companies, spokesperson Thomas Regnier confirmed on Sept. 1, 2026, at the Midday press briefing in Brussels. The AI Office's enforcement powers over providers of general-purpose AI, models that can be used for many different tasks, became applicable on Aug. 2. They cover transparency, copyright and safety risks from the most advanced models, and these requests are the first public sign that Brussels is using those powers in practice, not only preparing the framework.
Regnier described the steps as simple RFIs and the first enforcement steps, with dialogue continuing. Unlike a fine, an infringement decision or an order to pull a model, a request for information is a way to check whether a company is complying. The Commission refused to name recipients. Separately, Regnier confirmed very recent exchanges with OpenAI and Anthropic that included cybersecurity risks, but he would not say whether either company received an RFI. Those cyber contacts are not confirmation that either firm is on the RFI list.
Regnier said the RFIs fall in two main areas. The first covers safety and security, including general-purpose and the most advanced AI models. The second covers copyright and transparency and went to another set of AI companies. According to a news report on Sept. 2, Commission Vice-President Henna Virkkunen wrote that the copyright and transparency requests target companies that did not take part in informal compliance talks with the AI Office. The report also said the requests follow several AI model incidents this summer.
Under the Commission's published enforcement rules, the AI Office can send simple RFIs or formal ones issued by Commission decision. A company can be fined for an incorrect or misleading reply to a simple RFI; for a formal one, it can also be fined for failing to reply or replying incompletely. Breaches of the general-purpose AI rules can bring fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher, with higher ceilings for banned practices. The AI Office can also test models, ask for access to them and ask providers to take measures that can include restricting public access to a model.
A Commission press release on July 31, 2026, said enforcement of AI Act rules and new transparency requirements would begin Aug. 2, alongside a first list of more than 180 organizations signing the Code of Practice on transparency of AI-generated content. Regnier also said ENISA, the EU's cybersecurity agency, has been invited to be onboarded for access to advanced models, a process still under way in direct talks with the companies. The law takes effect in stages: rules for high-risk AI systems listed in its Annex III do not apply until Dec. 2, 2027.
It is not yet known which companies received which kind of request, whether the AI Office will follow up with model evaluations or formal RFIs, or whether this summer's incidents will lead to a formal finding against any company. The Commission has not published recipient names or an infringement decision, and reports naming specific labs as recipients remain unconfirmed.