It seems as if the EU legislators have managed to speed up discussions on the EU AI Act Omnibus law. This is the preliminary result, which still needs to be signed off by the Council of the European Union and the European Parliament. The planned changes largely reflect what we saw in the draft, but also contain some more restrictive new rules. In a nutshell: 

• Adds a new AI Act prohibition on generating non-consensual sexual or intimate content and child sexual abuse material.

• Sets delayed application dates for high-risk AI rules: 2 December 2027 for stand-alone high-risk AI systems and 2 August 2028 for high-risk AI systems embedded in products.

• Keeps the requirements that providers must register AI systems in the EU high-risk database when claiming an exemption from high-risk classification, and that processing special-category data for bias detection/correction must meet a strict necessity standard.

• Postpones national AI regulatory sandboxes to 2 August 2027 and shortens the grace period for transparency measures for artificially generated content, setting a new deadline of 2 December 2026.

• Clarifies the AI Office’s supervisory role over certain general-purpose AI-based systems, while preserving the competence of national authorities in areas such as law enforcement, border management, judicial authorities, and financial institutions.

• Creates mechanisms for industrial AI and sectoral legislation to avoid overlapping obligations, including exemptions and delegated powers under the Machinery Regulation, and requires Commission guidance to reduce compliance burdens.

• Next: Endorsement by the Council and European Parliament, followed by legal-linguistic review before formal adoption.

For additional information, read the full press release (published May 7 and updated May 18).