EU AI Omnibus: Parliament Votes — What the New Deadlines Mean for Your Business

The European Parliament voted today, 28 March 2026, in favour of its negotiating position on the AI Omnibus by 569 votes, formally opening the path to trilogue negotiations with the Council of the EU. The AI Omnibus forms part of the Commission\'s digital simplification package of November 2025, designed to reduce the administrative burden of the EU AI Act by at least 25% for all businesses — and by 35% for SMEs specifically.

Proposed Deadline Changes at a Glance

These proposals are not yet law. They represent Parliament\'s opening position for the trilogue and remain subject to negotiation with the Council:

High-risk AI systems (Annex III EU AI Act): Proposed delay of applicability to 2 December 2027. This covers systems involving biometrics, critical infrastructure, education, employment, essential services, law enforcement, justice and border management.

AI systems under EU sectoral safety legislation: Proposed delay to 2 August 2028 — giving providers of systems falling under existing product safety and market surveillance law significantly more preparation time.

Watermarking obligations for AI-generated content: Proposed deadline of 2 November 2026 — stricter than the Commission\'s own proposal of 2 February 2027, reflecting Parliament\'s more urgent concern about synthetic media.

AI regulatory sandboxes: Proposed delay in establishment deadline to 2 December 2027, aligned with the Council\'s position.

New: Proposed Ban on AI Nudifier Systems

Parliament has proposed an explicit prohibition on AI systems that generate or manipulate sexually explicit or intimate images of real, identifiable persons without their consent — so-called AI nudifiers. If agreed in trilogue, this would be added to the list of prohibited AI practices under Art. 5 EU AI Act. This is a parliamentary initiative, not a Commission proposal, and its survival in the negotiations remains to be seen.

Three Things Businesses Need to Know Now

First: 2 August 2026 remains the legally binding deadline until further notice. The proposed extensions are not yet in force. Businesses that pause compliance preparation in anticipation of a delay are taking a calculable legal risk — the trilogue could stall, collapse on procedural details, or produce a narrower outcome than expected.

Second: Political convergence between Parliament and Council is already visible. Both institutions support delaying high-risk AI deadlines and reducing administrative burdens. Outstanding divergence focuses on technical details — notably on AI literacy obligations (the Commission\'s proposal to shift responsibility to member states has faced pushback from both co-legislators) and the precise scope of the AI Office\'s supervisory mandate.

Third: The EU AI Office\'s role is being clarified. The Commission has confirmed the AI Office will oversee compliance of AI systems built on general-purpose AI (GPAI) models where the model provider and system developer are the same entity — with notable carve-outs for law enforcement, border management, judicial authorities, and certain financial institutions.

My Assessment

Today\'s vote confirms the political direction: the EU AI Act will be simplified, and the high-risk AI deadlines will almost certainly be extended. The remaining uncertainty is not whether, but by how much — and on what exact terms. For businesses currently building their AI Act compliance framework, I recommend a two-track approach: continue preparing for the core obligations under the current timeline, while structuring that preparation to absorb the likely deadline adjustments as they are confirmed in the coming weeks.

The first technical trilogue meetings are expected to begin today and continue through the Easter period. I will be tracking developments closely and will update this post as the negotiations progress.

The proposed deadline changes described above have not yet entered into force._