The European Union’s planned AI Act omnibus would postpone major compliance deadlines for high-risk systems, but it would not delay transparency requirements scheduled to apply on August 2, 2026. The distinction leaves providers of chatbots, generative AI tools and other systems that interact with people facing near-term obligations even as some high-risk developers receive substantially more time.
The omnibus reached political agreement on May 7, but its revised timetable will become legally effective only after formal adoption and publication in the EU’s Official Journal. Publication is expected before August 2. Until it occurs, however, the AI Act’s existing statutory calendar remains in place, creating uncertainty for companies preparing contracts, compliance statements and product documentation.
Once effective, the package would move the deadline for standalone high-risk systems covered by Annex III from August 2, 2026, to December 2, 2027. That category includes certain uses of AI in biometrics, critical infrastructure, education, employment, migration, asylum and border management. Requirements for high-risk systems incorporated into regulated products, including some machinery, lifts and toys, would be deferred until August 2, 2028.
Those extensions do not apply across the entire AI Act. Article 50 transparency rules remain due to take effect on August 2, 2026. They cover systems that communicate directly with people, generate or manipulate synthetic material, or perform functions such as emotion recognition. Depending on the system and its use, providers or deployers may need to tell people that they are interacting with AI or disclose that content was artificially generated or altered.
The omnibus provides a limited four-month transition for certain watermarking requirements involving systems already on the market. It does not amount to a general postponement for conversational agents, content-generation products or other systems subject to Article 50.
Separate duties for providers of general-purpose AI models have applied since August 2025 and are also unaffected by the high-risk extensions. Those rules include requirements involving technical documentation, copyright compliance policies and public summaries of model training content. The EU AI Office is responsible for important parts of the general-purpose AI regime, and the omnibus would expand its investigation and enforcement authority.
The AI Act’s prohibitions on specified unacceptable uses have applied since February 2025. The omnibus would add prohibitions related to AI-generated non-consensual intimate imagery and child sexual abuse material, accompanied by a transition period extending to December 2, 2026.
The gap between political agreement and publication matters because the revised high-risk dates are not yet part of applicable law. If the omnibus appears in the Official Journal before August 2, the postponements can take effect before the original high-risk deadline arrives. A delay in publication could briefly leave affected systems subject to the existing timetable, even if regulators are unlikely to prioritize enforcement based on a short procedural gap.
Private agreements may present a more immediate complication. Procurement terms, warranties and representations tied to “applicable law” can operate independently of regulators’ enforcement choices. Businesses relying on the postponed timetable therefore need to distinguish between deadlines already fixed in the AI Act and dates that remain conditional on publication of the omnibus.
The split schedule also makes accurate classification essential. A product using a foundation model is not necessarily an Annex III high-risk system, while a chatbot’s transparency duties do not disappear because separate rules for high-risk applications have been delayed. One service may also trigger different requirements depending on the provider, deployer, intended purpose and way it is offered in the EU.
Companies can confirm the legal change when the final measure receives a regulation number and appears through the Official Journal and EUR-Lex. Until then, the original high-risk timetable remains the formally published schedule, while the August 2 transparency deadline and existing general-purpose AI obligations continue independently of the omnibus.
Sources: EU AI Act