EU Parliament set to simplify AI Act rules and ban non-consensual nudification

The AI omnibus package is the EU’s attempt to make its landmark AI Act implementable without strangling innovation, while still showing it can react quickly to high‑profile abuses like deepfake nudification. By delaying some of the burdensome obligations on high‑risk systems and clarifying scope around research activities, Parliament is implicitly acknowledging industry complaints that the original timelines and definitions were too rigid for a technology moving this fast.

The targeted ban on non‑consensual nudification is symbolically powerful: it’s one of the first EU‑wide, use‑case‑specific prohibitions aimed squarely at a consumer‑facing generative AI harm. That sets a precedent for further content‑level bans if other abuses—say political deepfakes or automated fraud—become politically salient. For frontier labs, it means they must not only meet generic “high‑risk” requirements but also keep up with an expanding list of blacklisted applications.

From an AGI‑race perspective, the omnibus does not meaningfully slow capability work inside major labs. But it does harden the compliance environment for deployment in Europe, nudging companies to invest earlier in policy, logging and red‑team infrastructure if they want to ship cutting‑edge systems into the EU at scale.