Data · AI · Law · Ethics

The magazine for decision-makers

Digital Omnibus: what has already changed in the AI Act and what could change in the GDPR

The Digital Omnibus on AI is already law and shifts the AI Act deadlines. The one on data and the GDPR is still under discussion, and it is the more controversial of the two. A guide to what has been decided and what has not.

by

·

·

4–7 minutes
Il palazzo Berlaymont, sede della Commissione europea a Bruxelles

In brief

  • Already law: the Omnibus on AI is Regulation (EU) 2026/1744, in force since 27 July 2026. The obligations for high-risk AI systems are pushed back to 2 December 2027 and 2 August 2028.
  • Under discussion: the Omnibus on data and the GDPR is still being examined by the European Parliament. It touches the definition of personal data, legitimate interest for AI, cookies and data breaches.
  • For companies: more time under the AI Act does not mean everything can be postponed. Some obligations already apply.

On 19 November 2025 the European Commission presented the Digital Omnibus, a package to simplify the EU’s digital rules. In fact there were two proposals, with very different fates: one on artificial intelligence, approved within a few months, and one on data, the GDPR, ePrivacy and the Data Act, still under discussion and highly contested. It is worth keeping them apart.

The Omnibus on AI: an eight-month sprint

By European standards, the process was extremely fast. Parliament and Council reached agreement on 7 May 2026, the final plenary vote came on 16 June (423 in favor, 57 against, 174 abstentions), and publication in the Official Journal followed on 24 July.

OMNIBUS ON AI · FROM PROPOSAL TO LAW 19 Nov 2025Commissionproposal 13 Mar 2026Councilposition Mar 2026Parliamentposition 7 May 2026Trilogueagreement 16 Jun 2026Final voteParliament 24 Jul 2026Official JournalReg. 2026/1744 27 JulIn force Sources: European Parliament (Legislative Train, press releases), European Commission. Analysis by QrmServizi.
Eight months from proposal to the Official Journal.

The new AI Act deadlines

The Commission had proposed linking the application of the rules on high-risk systems to the availability of harmonized technical standards. The final text opted instead for fixed dates, which are easier to plan around:

AI ACT · THE NEW TIMETABLE 2 Aug 2026 Transparency Article 50 applies: chatbots disclosed, deepfakes labeled, content marked ALREADY IN FORCE 2 Dec 2026 New prohibition AI generating non-consensual sexual content or CSAM. Marking for systems already on the market FROM DECEMBER 2 Dec 2027 High risk Annex III: employment, credit, education, essential services and others +16 MONTHS 2 Aug 2028 High risk Annex I: AI embedded in regulated products (e.g. medical devices) +12 MONTHS Delays vs. original dates (2 Aug 2026 and 2 Aug 2027). Source: consolidated AI Act, Arts. 5, 50 and 113. Analysis: QrmServizi.
The key AI Act dates as they stand after Regulation (EU) 2026/1744.

Other important changes:

  • Ban on “nudification” apps. At Parliament’s urging, from 2 December 2026 systems that generate non-consensual intimate or sexual content or child sexual abuse material are prohibited.
  • AI literacy softened. Article 4 remains an obligation for providers and deployers, who must take measures to support AI literacy. But the text now clarifies that no specific level of competence has to be guaranteed, and the Commission and Member States must help, especially SMEs.
  • Simplifications for smaller companies. Simplified documentation is extended to medium-sized enterprises and so-called small mid-caps.
  • Sensitive data to correct bias. Processing it is permitted, but only where strictly necessary.
  • Machinery. AI embedded in machinery leaves the AI Act’s direct high-risk regime and is handled through the Machinery Regulation.

The Omnibus on data and the GDPR: still an open game

The second proposal (procedure 2025/0360(COD)) is still before the European Parliament’s ITRE and LIBE committees: the draft report dates from June 2026 and there has been no plenary vote yet. Here the changes go to the heart of the GDPR, which is why the debate is so heated.

GDPR · WHAT THE COMMISSION PROPOSES EDPB-EDPS OPINION “Relative” definition of personal datanot personal data for those who cannot identify the person OPPOSED Legitimate interest for developing and using AIexplicit legal basis for training SKEPTICAL “Residual” special-category data in AI datasetsexception for sensitive data present incidentally WITH CHANGES Cookies in the GDPR (Articles 88a and 88b)automated browser signals, no new request for 6 months after a refusal CAUTIOUS Data breaches: high risk only, 96 hourssingle notification point shared with NIS2 and DORA IN FAVOR Limits on abuse of the right of accessrefusal of requests used for purposes unrelated to privacy CAUTIOUS Source: proposal COM(2025) 837 and EDPB-EDPS Joint Opinion 2/2026 (11 February 2026). Summary by QrmServizi. Proposal still under discussion: the final text may change.
The most debated changes and how the European data protection authorities assess them.

The sticking point: the definition of personal data

This is the most sensitive change. Under the proposal, a piece of information would not necessarily be personal data for everyone merely because another party is able to identify the person. In practice, pseudonymized data could fall outside the scope of the GDPR for anyone who does not hold the key to re-identify it. The European Data Protection Board (EDPB) and the European Data Protection Supervisor (EDPS) have strongly urged lawmakers not to adopt this change.

The criticism

The noyb association, founded by Max Schrems, called the proposal “the biggest attack on Europeans’ digital rights in years”. The criticism concerns the subjective definition of personal data, the training of models on social network data, the limits on the right of access and the choice of a fast-track procedure without an impact assessment. On the other side, many companies have long been calling for simpler, more predictable rules, especially on cookies and breach notifications.

What to do now

  • Update your AI Act plan: the high-risk deadlines are further away, but transparency (Article 50), prohibitions and AI literacy already apply or will within a few weeks.
  • Do not anticipate a GDPR that does not yet exist: until the reform is approved, the current rules apply, including those on cookies and the 72-hour data breach notification.
  • Use the time gained: inventory your AI systems, classify their risk, document them. Those who reach 2027 prepared will have an advantage, not just one less compliance task.

The compass

Simplifying does not necessarily mean weakening, but the line is a fine one. On AI, Europe has chosen to give more time; on the GDPR, it has yet to decide how much data protection it is willing to trade for competitiveness.

This article is for informational purposes only and does not constitute legal advice. Information current as of November 2026.

Sources

Leave a Reply

Your email address will not be published. Required fields are marked *