What happened
prEN 18229-3, 'AI trustworthiness framework – Part 3: Human oversight' (AI Act standardisation request C(2025)3871 of 23 June 2025, CEN-CENELEC JTC 21 / WG 4, work item JT021056), reached the 'Under approval' stage on 22 October 2026, moving from the drafting stage. It has not yet been cited in the Official Journal, so it confers no presumption of conformity. Its next milestone is 1 July 2027; the AI Act Annex III high-risk obligations apply on 2 December 2027. The stage change was recorded from the public CEN-CENELEC work programme (tracked by the i6eal KI-Normungsmonitor / Standards Drift register, verified 26 October 2026).
Why it matters
Human oversight (AI Act Article 14) is one of the two explicitly security-relevant requirement areas in the Commission's standardisation request along with cybersecurity, robustness and accuracy, and it is the requirement that directly governs how autonomous/agentic systems are supervised. The fact that this draft is now 'under approval' signals that the second wave of AI Act harmonised standards (after EN 18286:2026, the quality-management standard already cited) is progressing. Providers and deployers of high-risk AI systems need these standards to achieve presumption of conformity; this one is still several steps from the Official Journal.
Action needed
Track the prEN 18229-series progress in the EU harmonised-standards pipeline; prepare gap analyses against the 'human oversight' requirements of Article 14 for high-risk systems, without relying on a presumption of conformity until the reference is published in the OJEU.