EU AI Act Digital Omnibus: Provisional Agreement Reached. Here Is What Changes

The European Parliament and the Council have reached a provisional agreement on the Digital Omnibus on AI proposal, which amends how and when parts of the EU AI Act apply.

Key changes

• Fixed timeline for high-risk rules: 2 December 2027 for stand-alone high-risk AI systems and 2 August 2028 for systems embedded in products.
• AI regulatory sandboxes: the deadline for Member States to establish at least one sandbox moves to 2 August 2027.
• Watermarking: obligations delayed until 2 December 2026.
• EU database registration: the obligation for providers to register high-risk AI systems is reinstated, along with the standard of “strict necessity” for processing special categories of personal data.
• Compliance support: the Commission is mandated to help operators of high-risk AI systems comply with minimal burden.
• AI Office: its competences are clarified.
• Transparency for AI-generated content: the grace period is cut from six months to three, with a new deadline of 2 December 2026.
• Overlaps with sectoral law: a mechanism lets the Commission, through implementing acts, limit certain AI Act requirements for industrial AI where similar AI-specific obligations already exist under sectoral legislation.

What happens next

The provisional agreement must now be formally adopted by the Council and the European Parliament.

Read the Council press release, and track changes like this one in the Asenion Global AI Regulation Tracker.

Originally shared on LinkedIn.