Javascript is required
logo-dastralogo-dastra

Digital Omnibus on AI : Parliament Votes, Deadlines Redrawn

Digital Omnibus on AI : Parliament Votes, Deadlines Redrawn
Leïla Sayssa
Leïla Sayssa
June 17, 2026·4 minutes read time

On 16 June 2026, the European Parliament cast its final vote approving the Digital Omnibus on AI, a package of targeted amendments to the EU AI Act. The Council is expected to formally adopt the text on 29 June 2026, after which it will be published in the Official Journal. For organisations operating in the EU AI space, the timetable has been restructured but not the framework itself.

Formal publication in the Official Journal is expected following Council adoption on 29 June 2026. We will continue to monitor and report on implementing developments.

A recalibrated timeline

The core architecture of the AI Act, which includes its risk-based classification, prohibited practices, and general-purpose AI rules, remains unchanged. Two categories of high-risk AI systems under the Act will now face extended application dates.

  • High-risk AI systems listed in Annex III and deployed on a standalone basis will be subject to full compliance requirements as of 2 December 2027, rather than 2 August 2026. That is a seventeen-month extension.
  • High-risk AI systems embedded as safety components in products regulated under existing EU safety legislation (Annex I) will not face those requirements until 2 August 2028 — twelve months later than the previous deadline of 2 August 2027.

For Article 50 transparency and watermarking obligations, the picture is more nuanced : AI systems already on the market before 2 August 2026 benefit from an extended deadline of 2 December 2026. For all other systems, the 2 August 2026 date continues to apply.

This distinction matters & renders the mapping of which systems are already deployed, and when, an important prerequisite for knowing which deadline actually applies.

Prohibition on AI-generated abuse content

Beyond calendar adjustments, the Omnibus introduces a substantive addition to the list of prohibited AI practices under Article 5.

AI systems designed to generate non-consensual intimate imagery (so-called nudification tools or child sexual abuse material) are now explicitly banned.

The prohibition covers both providers placing such systems on the EU market and deployers using them for those purposes. Compliance with this new prohibition is required by 2 December 2026.

Structural clarifications worth noting

The package also addresses several points of regulatory friction that practitioners have raised since the AI Act's adoption.

Overlapping obligations with existing sectoral safety legislation particularly in the machinery sector are being streamlined to avoid duplication. The concept of "safety component" receives further clarification, which will be relevant to product manufacturers navigating the boundary between product safety law and AI-specific requirements.

On data processing, the amendments permit the use of personal data, subject to appropriate safeguards, strictly for the purpose of detecting and correcting bias in AI systems.

SME-style flexibilities are extended to small mid-caps.

Enforcement of general-purpose AI model obligations is also being centralised at the level of the EU AI Office, which will have operational implications for how companies engage with regulators on GPAI matters.

Practical takeaway

The Omnibus buys meaningful time for organisations working to align high-risk AI systems with the Act's technical requirements; requirements that, in many cases, depend on harmonised standards not yet finalised. That rationale is legitimate.

However, 2 August 2026 remains a live and proximate compliance date for transparency obligations under Article 50.

The regulatory clock has not been stopped, rather it has been reset for some systems and left running for others.

For legal counsel, DPOs and compliance leads, the immediate priority remains the same: complete the inventory of AI systems in scope, determine which classification and which deadline applies to each, and sequence remediation accordingly.

For executive teams, the Omnibus offers breathing room on the most resource-intensive high-risk obligations, but it does not remove the need to have a documented AI governance posture in place well before the summer.


See Dastra in action

In just a few minutes, schedule a personalized demo and discover how Dastra can adapt to your organization.

Ask for a demo
Subscribe to our newsletter

We'll send you occasional emails to keep you informed about our latest news and updates to our solution

* You can unsubscribe at any time using the link provided in each newsletter.