
AI Omnibus
EU legislative package simplifying AI Act enforcement, splitting compliance into two deadline tracks.
Last refreshed: 17 May 2026
Does the AI Omnibus give European AI companies a real advantage, or just delay the compliance reckoning?
Timeline for AI Omnibus
Mentioned in: France chairs G7 Digital Ministerial on 29 May
European Tech SovereigntyDelayed Annex III compliance to 2 December 2027; GPAI deadline unchanged
European Tech Sovereignty: AI Omnibus deal splits enforcement into two speedsMentioned in: Mistral ships Le Chat Enterprise and Medium 3.5
European Tech SovereigntyMentioned in: Cohere-Aleph Alpha settle at 90/10, no filing yet
European Tech Sovereignty- What changed in the EU AI Act under the AI Omnibus deal?
- The 7 May 2026 agreement moved high-risk AI compliance from August 2026 to December 2027 (stand-alone) or August 2028 (embedded), left GPAI enforcement at August 2026, added an AI-generated CSAM/NCII prohibition, and eased compliance for SMEs.Source: Council of the EU, 7 May 2026
- When does the EU AI Act apply to high-risk AI systems after the Omnibus deal?
- After the AI Omnibus provisional agreement, stand-alone high-risk AI systems must comply by 2 December 2027, and high-risk AI embedded in products by 2 August 2028. GPAI rules remain at 2 August 2026.Source: Council of the EU, May 2026
- Why did the EU delay AI Act enforcement for high-risk systems?
- The EU delayed high-risk AI compliance deadlines in the AI Omnibus deal because the original August 2026 date was too soon for businesses to adapt, particularly SMEs. GPAI enforcement was left unchanged as large frontier-model providers had more time to prepare.Source: EU Consilium, May 2026
Background
The Digital Omnibus on AI (AI Omnibus) is an EU legislative package that amends the 2024 AI Act as part of the Commission's broader simplification agenda. On 7 May 2026, the Council of the EU and the European Parliament struck a provisional agreement that splits AI compliance into two speeds: Annex III high-risk AI system obligations were moved from 2 August 2026 to 2 December 2027 for stand-alone systems and to 2 August 2028 for high-risk AI embedded in products, while the General Purpose AI (GPAI) enforcement deadline was left unchanged at 2 August 2026.
The deal also added a new prohibition against AI-generated non-consensual intimate content and child sexual abuse material (CSAM), and extended simplified compliance documentation and lighter penalty treatment to small and medium enterprises and small mid-caps. The package forms part of the EU's Omnibus VII simplification legislative cycle and was treated as priority legislation given the August 2026 AI Act deadline that was otherwise imminent.
Formal adoption by Parliament and Council is required before 2 August 2026 to prevent the original high-risk deadline from triggering. The AI Omnibus is a bellwether for how the EU balances AI innovation policy with its regulatory apparatus; the two-speed approach is a concession to industry concerns without abandoning GPAI enforcement, which matters for large frontier-model developers including those operating from EU territory such as Mistral AI.