The AI Omnibus is Here: What the Parliament’s Vote Means for You

On 16 June, the European Parliament gave its final approval to the amendment of the AI Act under the AI Omnibus package. Formal adoption by the Council still remains, after which the text will be published in the Official Journal and enter into force. Here’s a breakdown of the biggest changes:

• More time for high-risk systems: The compliance deadlines have been pushed back. For Annex III systems, the new date is December 2, 2027, and for Annex I, it’s August 2, 2028. Keep in mind these dates are set in stone, they won’t move even if official guidelines are still being finalized.

• New bans: The list of prohibited AI practices now explicitly includes systems that generate non-consensual sexual content or child sexual abuse material (CSAM).

• Easier registration: If your system is exempt from high-risk status under Article 6(3), you still need to register it in the EU database, but the paperwork is now lighter. You’ll still need to document your internal assessment before going to market.

• Watermarking delay: The requirement to detect and watermark AI-generated content has been postponed until December 2, 2026.
• Bias correction: There’s now a clearer legal path to use sensitive data for fixing AI bias. It’s limited to what’s “strictly necessary” and requires strong safeguards, but it’s now available to more than just high-risk providers. This now covers a wider range of entities, not only providers of high-risk AI systems.

The core of the AI Act, like the risk-based rules, stays the same; the work is just being spread out. Now is the time for companies to start auditing and classifying their AI tools so they aren’t caught off guard later.

While these updates are a positive development, they do not fully meet the ambitions we advocated for at the AI Chamber. As detailed in our recent open letters [HERE and HERE], we called for deeper simplification and greater legal certainty to truly unlock the potential of AI development in Europe.