EU Seeks AI Act Answers From 30-Plus Companies, With Fines for Misleading Replies
The first information demands turn the EU’s general-purpose AI rules from written obligations into a live compliance process. A request is not a finding of wrongdoing, but the answers can determine whether the matter escalates.
Listen to this story
The audio brief
Story brief
3 key pointsThe EU AI Office has begun testing whether more than 30 global general-purpose AI providers can document compliance with the AI Act’s newly enforceable obligations. Companies must address model security, independent evaluations, post-market monitoring, and training-data disclosures. Inadequate or misleading replies can trigger fines of up to €15 million or 3% of worldwide annual turnover, plus possible corrective...
- 01
Requests seek evidence on model attack defenses, independent testing, post-market monitoring, and training-content summaries.
- 02
The Commission’s demands could precede formal investigations, but they are not findings of violations.
- 03
OpenAI, Anthropic, and Google were reportedly contacted; the Commission has not published a recipient list.
The European Commission has sent information requests to more than 30 AI companies worldwide under the EU AI Act, opening a preliminary process that can lead to formal investigations. The requests put providers’ safety and copyright practices under direct regulatory scrutiny.
Henna Virkkunen, the Commission’s executive vice-president for tech sovereignty, said on August 29 that the EU AI Office had formally contacted providers of general-purpose AI models in different regions. Those obligations became enforceable on August 2, and the Act’s transparency rules took effect in August.
The records now under review
- How providers secure models against attack.
- Whether they have conducted independent external evaluations.
- How they monitor models after they enter the market.
- Detailed training-content summaries from providers that have not published them or joined the AI Office’s informal compliance dialogues.
The wider law prohibits some AI activities judged to carry unacceptable risk and sets transparency standards for other AI services. The new requests narrow that broad framework into specific materials the Commission can assess, rather than an immediate ruling that a provider has violated the Act.
The Commission said the requests went to more than 30 AI companies worldwide and may precede formal investigations.
Incorrect, incomplete or misleading responses may bring a fine of up to €15 million 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. No such action has been announced. Euractiv reportedly identified OpenAI, Anthropic and Google among the recipients, though the supplied evidence does not provide a Commission-published recipient list.
Washington argues for a different rulebook
At a G20 innovation ministerial meeting, the United States argued against AI-specific rules and promoted the technology-neutral Carolina Principles. US tech adviser Michael Kratsios said policymakers should not treat each emerging technology as a separate policy problem. Mark Zuckerberg and Elon Musk also argued for fewer constraints on AI expansion.
That contrast leaves companies facing two distinct policy approaches: Washington’s call to avoid AI-specific regulation, and Brussels’s first formal demands for evidence on model safeguards and transparency. The immediate unresolved question is whether the Commission finds answers that warrant an investigation or corrective action.
Editorial analysis
Our Read
Our Read: The important shift is procedural. The AI Office is building a record on how providers secure, test and monitor models after release before it decides whether to pursue investigations or corrective measures. The next useful signal is whether any request produces a formal case, not the request alone. OpenAI may also face a separate European compliance track: ChatGPT was recently designated as an online search engine under the Digital Services Act, with its own risk-management duties. That does not establish the same overlap for other providers.
Sources
- aljazeera.comUS pushes looser approach to AI regulation, while EU pushes new law
- tokenstead.aiEU AI Act Enforcement Begins: The AI Office Starts Asking
