In March 2026 the European Parliament adopted its position on the simplification (omnibus) proposal amending the AI Act. The proposal originates from the seventh simplification package tabled by the European Commission on 19 November 2025 — the “digital omnibus”.
The problem the proposal addresses
The Commission’s stated rationale is concrete: standards for high-risk AI systems have been delayed, and the designation of national competent authorities and conformity assessment bodies has been slow. Applying the original timetable as-is would leave companies facing a gap in which “the rules are in force but the supporting infrastructure is not ready”.
The core move is therefore to push back the application of certain high-risk AI rules, so that standards and bodies can be in place before substantive compliance begins.
The Parliament’s position adds a call to ban “nudifier” apps — services that use generative technology to produce nude images of real people, one of the most closely watched deepfake scenarios.
What it means for Chinese companies going global
First, the timetable is reordered, not simply postponed. Delaying high-risk rules does not mean compliance work can be delayed. Technical documentation, risk management, data governance and log retention are required by every version of the rules — see our EU AI Act service.
Second, transparency duties are not part of the delay. A distinction must be drawn between “rules that are postponed” and “rules already in force”. The Article 50 transparency and labelling duties apply from 2 August 2026, and their direction is settled — see our note on the EU AI Act transparency rules taking effect.
Third, prohibitions carry greater certainty. The stance against deepfake pornography has been consistently reinforced at Parliament level — see our note on the EU agreement to ban deepfake pornography and simplify high-risk rules. For products involving human-likeness generation, these red lines belong in the design stage, not in post-launch filtering.
What companies should do now
First, build a “rule — timetable — owner” mapping table. EU rules change frequently, and one table tracking the application date and owner of each duty is more reliable than scattered monitoring.
Second, make the high-risk classification judgement real. Classification determines which set of rules applies and is where companies most often misjudge. For a multi-jurisdiction approach, see our multi-jurisdiction service.
Third, compare China, the US and the EU. The EU is simplifying, the US is advancing a federal framework, and China is advancing legislation and filing — the rhythms differ, and companies need a baseline that satisfies all three. See our note on the Manus foreign-investment security review.
The EU’s “simplification” is not deregulation but reordering: put the supporting infrastructure in place first, then apply. For companies, reading the sequence matters more than reading the clauses.
This article is based on public European Parliament documents and Commission proposal information from March 2026. It is for general reference only and does not constitute legal advice.