What happened. On 7 May 2026, the EU Council and Parliament reached a political agreement on the so-called “AI Omnibus” — a package of targeted amendments to the EU AI Act proposed by the Commission in November 2025. The agreement marks the end of months of interinstitutional negotiations and provides clarity on several practical disputed questions.
Context
The AI Omnibus was announced as a “simplification package”, which in Brussels typically means “relaxation of certain obligations plus tightening in other areas”. The 7 May agreement confirms this expectation. First, the deadlines for several obligations are being staggered — industry had argued that the original deadlines did not align with the availability of concrete guidelines. This argument has been partially upheld: certain sector-specific obligations are being deferred by six to twelve months.
Second, several definitions have been sharpened, particularly the distinction between “provider” and “deployer”. In practice, this distinction was especially problematic in reseller models — whoever distributes an OpenAI model under their own label was sometimes treated as one, sometimes as the other. The new version provides greater clarity here.
Third — and this is where EDPB and EDPS raised the loudest objections — certain high-risk thresholds are being changed. Data protection authorities warn that this reduces the level of protection for data subjects without the EU Charter being subjected to a cross-examination. This criticism has been softened in the final text, but not fully addressed.
What this means in practice
For the majority of companies deploying AI, the most important changes are the sharpened definitions and deferred deadlines. Those who already have an EU AI Act compliance plan should review it on two points: which obligations were tied to unclear thresholds that are now more sharply defined? And which deadlines were tight in their own planning and are now relaxed?
Those without a plan should not misunderstand the Omnibus as a deferral. The core obligations — risk classification, transparency, high-risk requirements — remain. Only the interpretation becomes clearer, not milder.
Original Sources
- EU Council — Press Release of 7 May 2026
- EDPB & EDPS — Joint Opinion on Digital Omnibus on AI
- EU Commission — Regulatory Framework AI (as of 2026)
— Lumi AI Act Watch · 25 May 2026. Research and first draft AI-assisted, editorial approval by Lumi-Systems.io. Disclosure in accordance with Art. 50 EU AI Act.