The European Commission published two draft Regulations on November 19, 2025: the Digital Omnibus (available here) and the Digital Omnibus on AI (available here).

The European Commission launched a Digital Package on Simplification as part of its broader initiative to simplify the EU rulebook. Serving as a stress test for existing digital rules, Digital Omnibus aims at addressing stakeholders’ concerns with an ambitious list of amendments.

The draft Digital Omnibus on AI mainly aims to simplify the implementation of harmonised rules on artificial intelligence as provided for in Regulations (EU) 2024/1689 on artificial intelligence (so called “AI Act”). This article highlights the key amendments proposed by the European Commission in this regard.

Simplification for SMES and SMCS

The Digital Omnibus on AI broadens the proportionality measures that the AI Act originally granted only to small and medium-sized enterprises. These simplifications would also apply to small mid-cap companies (SMCs),i.e. enterprises larger than SMEs, which employ fewer than 750 employees, and have an annual turnover not exceeding EUR 150 million or an annual balance sheet total not exceeding EUR 129 million.

The simplifications include the possibility to receive targeted guidance and support, from competent national authorities. Moreover, when penalties apply, Member States must consider SMEs’ and SMCs’ economic viability when determining sanctions.

Simplification for high-risk system

General simplifications. The current AI Act requires all high-risk systems to be registered in a European database. This obligation would be removed for systems that are exempted from high-risk classification because they are for instance used only for preparatory tasks.

The proposal also clarifies how the conformity assessment procedure applies when a high-risk AI system is subject both to the AI Act and to Union product legislation.

Specific simplifications for SMEs and SMCs. SMEs and SMCs developing high-risk AI systems, including start-ups, would be able to submit simplified technical documentation, and to apply a simplified Quality Management System (QMS).

Expended scope of regulatory sandboxes and Real-World Testing

AI regulatory sandboxes, which are controlled environments where innovators can test AI technologies with guidance and supervision from authorities, would be expanded. The new regulation gives the European Commission’s AI Office a formal legal basis to create an EU-level sandbox for systems falling under its competence, for example, general-purpose AI (GPAI) or AI integrated in very large platforms). National sandboxes would also be required to facilitate cross-border cooperation.

Real-world testing framework is expended to include high-risk AI systems integrated into products governed by EU product safety legislation, allowing such systems to be tested in real-world conditions outside regulatory sandboxes. For stand-alone high-risk AI systems, the Digital Omnibus on AI introduces a legal basis for voluntary cooperation agreements between interested Member States and the European Commission to authorize and coordinate real-world testing.

Data protection and cooperation with fundamental rights authorities

The Digital Omnibus on AI establishes a specific legal basis allowing the processing of special categories of personal data (such as data revealing racial or ethnic origin, health or sexual orientation) when this is strictly necessary to detect and correct bias in AI systems.

Moreover, cooperation between market surveillance authorities and fundamental rights bodies is reinforced. These authorities would have reciprocal access to relevant information and an explicit mutual obligation to cooperate.

Other simplification measures

Non-binding nature of codes of practice. The Digital Omnibus removes the Commission’s power to approve Codes of practice (including those on transparency and GPAI), which would now remain voluntary guidance instruments coordinated by the AI Office. The Commission may, however, adopt binding common rules if these voluntary codes prove insufficient to ensure compliance.

AI literacy. The previous mandatory requirement for providers and deployers to train their staff is replaced by a general obligation for the Commission and Member States to encourage and promote AI literacy initiatives rather than impose it directly.

Post-market monitoring. Instead of a uniform EU monitoring template, companies would be allowed to develop their own monitoring plans, guided by Commission recommendations.

Governance and supervision

AI Office would have exclusive competence for GPAI systems, where the same company develops both the underlying model and the deployed system, and for AI systems integrated into very large online platforms or search engines (VLOPs/VLSEs).

The Digital Omnibus on AI also harmonizes and simplifies the designation process for conformity assessment bodies by introducing a single application and assessment procedure, aligning classifications within the NANDO information system.

Implementation timelines and transitional measures

High risks AI systems. Obligations for most high-risk AI systems would apply six months after the Commission confirms that the measures in support of compliance with the AI Act’s are available. For systems subject to EU product-safety legislation, the obligations would apply twelve months after the same decision.

In any event, all high-risk obligations must apply by December 2027 for most high-risk AI systems and by August 2028 for high-risk systems subject to EU product-safety legislation.

Transparency obligations. Regarding transparency obligations, a six-month grace period would apply for providers of generative AI systems already on the market before August 2026 to comply with new labelling requirements for artificially generated or manipulated content.

Next steps

Open consultation. The draft Digital Omnibus on AIis open for feedback for a period of 8 week(s) from the day this adopted proposal is available in all EU languages – so far until end of January 2026.

Official discussions. The draft Digital Omnibus on AI will be submitted to the European Parliament and the Council for adoption. The draft Regulation will be examined by Members of the European Parliament from January 2026 with the aim of adopting a final report by Q1 2026. In parallel, Member States will begin discussions within the Council to prepare their position. From Q2 2026, the European Commission, European Parliament, and Council would initiate “trialogue” negotiations to reach a compromise text. It is likely that the Parliament will apply urgent procedure to bypass Committee negotiations and go directly with a vote of the text, thus reducing opportunities to present amendments. In this case, the adoption of the text could be expected by Q1 2026 instead of mid-2026.