A European Commission draft would let investigations into certain AI systems reset the five-year clock for imposing penalties, including when the Commission takes specified investigative steps. It also proposes rules for what companies under investigation, and their outside advisers, can see in the case file.
Watch Desk analysis
What happened
The draft concerns proceedings under Articles 75c and 75d of the EU AI Act. Its scope includes some AI systems built into designated very large online platforms or search engines, and systems developed alongside their underlying general-purpose AI models by the same provider or undertaking. The draft names no companies.
For continuing or repeated non-compliance, the five-year period would begin when the non-compliance ends. Requests for information, inspections and the opening or reopening of proceedings are among the listed actions that would interrupt the period and start it running afresh. The Commission draft also sets out how investigated operators and specified external legal, economic and technical advisers could access case documents. The proposal is not adopted law: the Commission describes it as preliminary. PPC Land reports that the consultation runs until 9 November 2026, with adoption planned for the fourth quarter. Read PPC Land’s account of the draft.
Why it matters
The proposed clock is not simply five years from the first suspected breach. For an ongoing breach, the clock would not start until it stops, and listed investigative actions could restart the period. That could give the Commission more time to pursue cases, while leaving companies and other parties with questions about how evidence and confidential material would be handled.
The file-access rules also draw a line between the company and its advisers: the draft describes access to unredacted documents for a limited number of specified external advisers under disclosure terms, rather than unrestricted access for the company itself. That is a consequential procedural choice, not just Brussels paperwork with a very long appendix.
Our read
This is a meaningful proposed rulebook for how certain AI enforcement cases would work: when the penalty clock runs, what can reset it, and who can examine the evidence. But it remains a draft, so the consultation is the moment for affected companies and other interested parties to contest the balance between effective enforcement, confidentiality and the right to answer a case.
What to watch
- Whether the final text keeps the listed actions that restart the limitation period.
- How the Commission responds to feedback on access to confidential evidence.
- Whether the planned fourth-quarter adoption timetable holds.
Discussion spark: Should investigative steps restart the five-year fine clock, or should regulators have a fixed deadline once a suspected breach is known?
Sources and evidence
- EU draft lets AI Office restart 5-year fining clock on Google and Meta AI – PPC Land (12 October 2026, 16:03 UTC)
- GE Aerospace's AI Defense Opportunity With Shield AI – GE Aerospace (NYSE:GE) – Benzinga (12 October 2026, 16:03 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.