AI Omnibus takes effect with revised dates for high-risk systems

The AI Omnibus entered into force across the EU on 27 July 2026, according to the Commission. It changes compliance timelines, administrative obligations and oversight provisions under the AI rulebook. The measure also prohibits specified AI-generated sexual content and expands supervised testing opportunities.

The Commission describes the Omnibus as a targeted simplification proposed on 19 November 2025 within a digital omnibus package. Entry into force marks the measure's legal start; it does not mean every revised rule applies on that date. The published application dates for high-risk categories extend into later years. The text presents the changes alongside safeguards for safety and fundamental rights.

Which dates govern the high-risk rules?

The Commission states that rules for high-risk AI systems in Annex III apply starting 2 December 2027. A distinct timeline applies to high-risk AI embedded in physical products covered by Annex I: those rules apply starting 2 August 2028. The source lists machinery, toys and lifts as examples of physical products. These dates describe application of the rules, rather than the Omnibus's entry into force.

What administrative changes are specified?

The measure extends some provisions previously reserved for SMEs to small and mid-cap companies, referred to as SMCs. The source describes simplified legal obligations for SMCs but does not set out eligibility tests or company-specific outcomes. It also simplifies the previous AI literacy requirement for companies. The Commission and Member States are assigned a stronger role in promoting AI literacy.

Registration of exempted AI systems in the EU central database is also simplified. The source does not give the revised registration steps or identify particular systems affected. In addition, it says the Omnibus clarifies the relationship between the AI Act and other EU laws. Procedures for conformity assessment bodies are also simplified, without further procedural detail in the source.

What oversight and safeguards change?

The AI Office receives extended oversight of certain AI systems. The stated scope includes systems built on general-purpose models and systems embedded in large online platforms and search engines. The source does not identify individual systems or set out case-specific enforcement actions. It characterises the change as an extension of governance powers.

The Omnibus prohibits AI systems that generate non-consensual sexually explicit and intimate content or child sexual abuse material. It also allows processing special categories of personal data for detecting and correcting bias. The source describes these as safety and fundamental-rights provisions. It does not report an individual company response or an enforcement decision concerning a named company.

The measure expands access to regulatory sandboxes and introduces an EU-level regulatory sandbox. These are supervised settings for developing and testing AI systems, according to the Commission's description. The source does not provide operating procedures or name participating companies. The legislative text and implementation timeline are identified as further reference materials.