Data, AI & Digital Sovereignty
9. EU AI Act Omnibus reaches political agreement — high-risk obligations delayed to December 2027 and August 2028
On 7 May 2026, EU co-legislators reached a provisional political agreement on the Digital Omnibus on AI, having overcome a stalled trilogue that collapsed on 28 April over conformity-assessment architecture for Annex I embedded systems.
BY GEOPOLITICS DESK · JUNE 4, 2026 · 1 MIN READ
On 7 May 2026, EU co-legislators reached a provisional political agreement on the Digital Omnibus on AI, having overcome a stalled trilogue that collapsed on 28 April over conformity-assessment architecture for Annex I embedded systems. The agreement delays the application of high-risk AI Act obligations: Annex III systems to 2 December 2027; Annex I embedded systems to 2 August 2028. Generative-AI content-marking obligations under Article 50 remain on track for 2 August 2026 (with a transitional period for pre-market systems). The AI Omnibus does not remove core obligations — it recalibrates timelines and reduces governance fragmentation through a centralized AI Office oversight mechanism. In-house legal teams deploying AI systems in regulated sectors should update compliance roadmaps to reflect the new application dates and ensure their AI Office registration and transparency timelines are current.
| Bird & Bird — Digital Omnibus on AI Trilogue