AI Act: the EU revises its own timeline with the ‘Digital Omnibus’

The Digital Omnibus on AI entered into force on 27 July 2026, amending select implementation deadlines of the EU AI Act (Regulation (EU) 2024/1689) and simplifying part of the compliance regime, while introducing new protections against non-consensual intimate deepfakes. Article 50 transparency obligations remain applicable from 2 August 2026, as originally scheduled.

This is the first substantive revision of the text since it entered into force on 1 August 2024. The original calendar phased in obligations step by step: a ban on prohibited practices in February 2025, general-purpose AI model obligations in August 2025, high-risk system rules (Annex III) and transparency duties in August 2026, and regulated products (Annex I) in August 2027.

For the sectors Cardan-AI tracks — aerospace, defence, energy, luxury — the exact scope of the Omnibus still needs to be worked through instrument by instrument, but the signal is clear: Brussels is adjusting its own calendar in response to implementation friction reported by companies and member states, just two years after the regulation took effect.

For Cardan-AI, the episode illustrates a point in the economics of regulation: a regulatory calendar that gets revised mid-course becomes itself a source of uncertainty that firms must price into compliance investment decisions — regardless of whether the revision tightens or loosens the rules. Full analysis at cardan-ai.com.