EU AI Act: Omnibus trilogue negotiations fail - high-risk obligations risk taking effect in August 2026 without delay
What it really says
The second political trilogue between the European Parliament, the Council of the EU, and the European Commission on the Digital AI Omnibus ended on April 28, 2026, after roughly 12 hours of negotiations without agreement. The main sticking point is Annex I: the conformity assessment architecture for AI systems embedded in regulated products such as industrial machinery, medical devices, and in-vitro diagnostics. Parliament wants to shift conformity assessment for certain product categories from Section A to Section B, moving to sectoral assessment procedures. The Council objects. This single disagreement blocked the entire package. The Omnibus's most significant change would defer the AI Act's high-risk obligations: from the original date of August 2, 2026, to December 2, 2027, for standalone high-risk AI systems under Annex III (including AI in employment contexts), and to August 2, 2028, for AI embedded in regulated products under Annex I. Without timely agreement, the original obligations take effect on August 2, 2026, as written. The next trilogue is scheduled for May 13, 2026. The Cypriot Council Presidency must reach a result by June 30, 2026, after which Lithuania takes over.
Our assessment
The situation is concerning for businesses, but context shows this is not a refusal to regulate AI - it's about fine-tuning implementation. The EU is regulating AI; the question is when specific obligations take effect. For companies developing or deploying AI systems, the impasse creates genuine uncertainty. If the Omnibus isn't passed in time, high-risk obligations could apply from August 2026 - obligations many companies aren't prepared for. At the same time, critics argue the proposed 16-month delay means potentially risky AI systems operate longer without oversight - such as AI in hiring processes or medical diagnostics. The truth lies between: a rushed, poorly implemented regulation helps no one. But a delay misread as a free pass damages trust in the technology.
Relevance for Germany
Directly relevant for Germany. German companies must prepare for two scenarios: either high-risk obligations take effect on August 2, 2026 - leaving only three months - or the Omnibus passes and pushes the deadline to late 2027. Given the uncertain outcome, legal experts recommend aligning compliance preparations with the original timeline. Particularly affected are companies using AI in human resources (applicant screening, performance evaluation), medical diagnostics, critical infrastructure, and public administration. Germany's IHK and Bitkom have already noted that many German SMEs are not yet AI Act-compliant. The failed trilogue increases pressure. The next negotiation round on May 13 will be decisive - time is running short for the Cypriot Presidency.
Fact check
Core facts are well-documented by legal specialist sources. The failed trilogue on April 28 after 12 hours is consistently reported by IAPP, Ropes & Gray, and Computerworld. The Annex I conformity assessment dispute is analyzed in detail by A&O Shearman and Ropes & Gray. The proposed new deadlines (December 2, 2027, for Annex III; August 2, 2028, for Annex I) come from DLA Piper's analysis of the Omnibus draft. The next trilogue date of May 13 is confirmed by multiple sources. That original deadlines apply automatically without agreement is codified in the AI Act (Regulation 2024/1689, Article 113).
Source
- • IAPP 02.05.2026 (iapp.org/news/a/ai-act-omnibus-what-just-happened-and-what-comes-next)
- • Ropes & Gray April 2026 (ropesgray.com/en/insights/viewpoints/102mquz/ai-omnibus-trilogue-underwaywhat-to-expect-as-negotiations-progress)
- • A&O Shearman April 2026 (aoshearman.com/en/insights/digital-omnibus-on-ai-what-is-really-on-the-table-as-trilogues-begin)
- • DLA Piper 2026 (knowledge.dlapiper.com - The Digital AI Omnibus: Proposed deferral of high risk AI obligations)
- • Computerworld May 2026 (computerworld.com/article/4164963/eu-lawmakers-fail-to-agree-on-watered-down-ai-act)