Brussels recalibrates the AI Act: what the Digital Omnibus changes

The EU is giving itself more time to apply the heaviest obligations of the AI Act. On 7 May 2026 the European Council and the European Parliament reached a provisional agreement on the Digital Omnibus on AI, and Parliament voted to adopt it on 16 June 2026; formal Council approval was still required, and both institutions aimed to complete adoption before 2 August 2026. Under the agreement, high-risk obligations for stand-alone systems listed in Annex III, such as those used in employment, would apply from 2 December 2027, and for AI embedded in regulated products such as medical devices from 2 August 2028. Some rules move faster: a ban on AI systems generating child sexual abuse material or non-consensual intimate content, and a watermarking obligation for AI-generated content for systems already on the market, both from 2 December 2026. The EU AI Office will supervise AI systems built on general-purpose models from the same provider. Why it matters: the world's first comprehensive AI law is being recalibrated between ambition and implementation capacity. Source: Sidley Austin, Data Matters, 22 June 2026.