Luxembourg: The Digital Omnibus on AI, published as Regulation (EU) 2026/1744, has been in force since 27 July 2026 and has moved the main compliance dates for high-risk artificial intelligence systems to December 2027 and August 2028. Providers now have one month left before the next date that still bites, 2 December 2026.

The Digital Omnibus on AI amends the AI Act, the European Union’s flagship law on artificial intelligence. It entered into force three days after publication in the Official Journal, according to a Hunton legal briefing from 28 July 2026. The European Commission proposed the package in November 2025, citing late designation of national authorities and conformity assessment bodies, and the lack of harmonised standards and guidance for high-risk systems.

The new calendar splits the obligations in four steps. Transparency obligations began to apply on 2 August 2026. Certain rules on marking AI-generated content will reach systems that were already on the market before that date on 2 December 2026. The core duties for stand-alone high-risk systems listed in Annex III, which include tools used in employment and education, now apply from 2 December 2027. High-risk AI embedded in regulated products under Annex I follows on 2 August 2028.

For companies, the 2 December 2026 date is the practical deadline this winter. Any provider whose chatbot, image generator or synthetic media tool entered the market before 2 August must now prepare to mark its outputs. Compliance teams should check whether their current systems already label content in a machine-readable way, because the Digital Omnibus on AI gives no extra grace period beyond that date.

The package also adds a new prohibition. It bans AI systems designed to generate non-consensual sexual or intimate imagery and child sexual abuse material. Lawmakers added the ban while they were simplifying the rulebook, which shows that the Digital Omnibus on AI does not only loosen obligations. It tightens the list of banned practices at the same time.

Business gets some relief too. The Digital Omnibus on AI extends certain simplified documentation measures to medium-sized companies, and it reduces some administrative burdens. It streamlines conformity assessment and reinstates a simplified registration requirement for systems that providers consider not to be high risk. At the same time it restores a stricter standard for processing special-category data to detect and correct bias, so developers cannot treat the omnibus as a blanket deregulation.

The AI Office gains weight as well. The text expands its powers over general-purpose AI models, which concentrates oversight of the biggest model providers in Brussels instead of leaving it entirely to national regulators. National authorities remain responsible for most high-risk systems, so the pace of their designation will decide how smoothly the delay turns into real enforcement in 2027.

The delay is the main political story. Critics of the AI Act argued for months that Europe was regulating faster than its standards bodies could write the technical rules companies need. The Digital Omnibus on AI answers that complaint by giving standardisation roughly sixteen more months for Annex III systems. Civil society groups will watch whether the extra time produces usable standards or simply postpones the hard choices.

Employers and schools deserve particular attention. Recruitment software, worker monitoring tools and systems that score students sit in Annex III, so the December 2027 date decides when human oversight, data governance and logging duties become binding for them. Until then, general EU law on data protection and equal treatment continues to apply to these tools.

The next milestone is close. Providers of generative AI should confirm before 2 December 2026 that their marking solutions work, and compliance officers should map every system against the new Annex III and Annex I dates. The Digital Omnibus on AI has bought the industry time, but it has not removed the obligations.