EU AI Omnibus Final Approval: High-Risk Deadline Extended — August 2 Transparency Obligations Still Live

The Council of the EU gave final approval on July 9 to the Digital Omnibus on AI, the regulation updating the EU AI Act's timeline and technical requirements. The headline change is that Annex III stand-alone high-risk AI system obligations (Article 6(2)) have been moved from August 2, 2026 to December 2, 2027 — a 16-month extension — while high-risk AI embedded in regulated products (Article 6(1), Annex I) has been pushed to August 2, 2028. What remains live from August 2, 2026: Article 50 transparency obligations (AI-generated content must be disclosed and labeled; systems interacting directly with humans must identify themselves as AI); GPAI model enforcement powers (the AI Office can now request information, require fixes, and act on non-compliance); and the Article 5 prohibited-practice rules, which have applied since February 2025. A further change from December 2, 2026: Article 5 bans two new categories — systems that generate realistic intimate images of identifiable persons without explicit consent, and systems that generate child sexual abuse material. For innovation leads and general counsel, the Omnibus creates a coherent short-term compliance checklist: transparency labeling for AI-generated content is the immediate priority; high-risk system conformity assessment programs remain necessary but have an extended runway; and registration and Article 6(3) self-assessments must remain on the compliance calendar.