EU Commission drafts rules for fining ChatGPT, Gemini under AI Act

Italian site pasqualepillitteri.it reports that on October 12, 2026 the European Commission put out a draft implementing regulation laying out how the AI Office will run proceedings that can end in AI Act fines for systems it supervises directly, ChatGPT and Gemini among them. The draft, a ten-page text with nine articles filed as Ares(2026)9589403, covers access to the file, trade secrets and limitation periods. The public can comment through November 9, 2026, until the end of that day in Brussels, using a free EU Login account.

The AI Office directly oversees applications built on a general-purpose model from the same company, and AI systems that are or sit inside very large online platforms or search engines designated under the Digital Services Act. The article lists exceptions, including systems linked to products regulated elsewhere, critical infrastructure, law enforcement, border control, financial institutions and the administration of justice.

On access to the file, a company can see the AI Office's documents only after receiving preliminary findings. Third parties' trade secrets are blacked out, and unredacted versions go to a small data room open to named outside advisers who sign disclosure terms. Under Article 3, anyone submitting documents may be asked to justify confidentiality line by line, and unjustified claims lose protection.

The Office has five years to impose a fine, counted from the violation or its end. Each investigative act restarts the clock, subject to a ten-year cap, and collecting a final fine allows five more years. The article says the draft does not alter fine amounts, which already follow Article 99, with the lower of the two caps applying to small and medium-sized firms.