What happened
On 12 October 2026 the European Commission opened a four-week consultation (closing 9 November 2026) on a 10-page draft Implementing Regulation (Ares(2026)9589403) setting out proceedings under new AI Act Articles 75c and 75d (inserted by the Digital Omnibus on AI, Regulation (EU) 2026/1744). The draft rulebook covers access to the file, procedural safeguards/confidentiality, and limitation periods for penalties for the AI systems the AI Office exclusively supervises: GP-model-integrated AI in very large online platforms/search engines and in-house general-purpose AI models. Notably, each of six investigative steps can restart the five-year limitation clock (up to a 10-year cap). Adoption is planned for Q4 2026.
Why it matters
This is the operational rulebook for the AI Office's new exclusive enforcement power over AI features inside the largest platforms — the first substantive codification of how fines (up to 7% global turnover), periodic penalty payments (5% daily) and defence rights will actually run. It defines procedural fairness, business-secret protection and how long the Commission has to fine Google/Meta-scale operators, materially shaping enforcement risk for any company whose AI is integrated into a designated VLOP/VLOSE or built on its own GP model.
Action needed
Operators in scope (VLOP/VLOSE AI features, in-house model-and-system providers) should submit feedback by 9 November 2026 and prepare an access-to-file / negotiation strategy for future AI Office proceedings; anticipate a restarted 5-year fining clock during AI Office investigations.