Digital Omnibus on AI: Greater Flexibility, but Growing Uncertainty for AI Act Compliance
Tuesday, 28 January 2025On 19 November 2025, the European Commission published its Digital Omnibus on AI Regulation Proposal (the “Proposal”), introducing targeted amendments to the EU AI Act as part of a broader drive to simplify EU digital regulation in the face of push back from stakeholders . While the Proposal is intended to ease compliance burdens, particularly in respect high risk AI systems and for entities which come under the definition of “Small Mid-Caps”, it also creates significant uncertainty around implementation timelines.
We have seen first hand the issues that the Proposal has presented to our clients in this space.
Conditional Timelines under the AI Act
A central feature of the Digital Omnibus is the proposal to decouple the application of key AI Act obligations from dates set out in the Regulation currently and, instead, link them to the future availability of Commission guidelines, harmonised standards and other compliance support tools.
Paragraph 31 of the Proposal amends Article 113 of the Regulation, which sets out when the EU AI Act will apply. Under the Proposal, Chapters III, Sections 1 to 3 (covering the classification of high risk AI systems, the requirements for such systems, and the obligations of providers, deployers and other parties) will apply only once the Commission has adopted a decision confirming that adequate compliance support measures are in place. From the date of that decision, these provisions will apply:
- after 6 months for AI systems classified as high risk under Article 6(2) and Annex III; and
- after 12 months for AI systems classified as high risk under Article 6(1) and Annex I.
The Proposal also introduces long stop dates. For AI systems classified as high risk under Article 6(2) and Annex III, the relevant provisions will apply from 2 December 2027, and for AI systems classified as high risk under Article 6(1) and Annex I, from 2 August 2028. From these dates, the High Risk AI provisions will apply regardless of whether the compliance support measures are in place. The effect of the above Proposal will be to mean that at any time between August 2026 and December 2027, Companies will be told that they have 6 months within which to comply.
A Regulatory “Limbo”
For many organisations, this approach creates a regulatory limbo. Clients designing AI governance frameworks and compliance roadmaps are now required to prepare for obligations that are in place under legislation that may be amended at an unspecified time.
This uncertainty presents practical challenges:
- Compliance programmes risk being over or under engineered in the absence of final technical standards and guidance;
- Long lead AI development and procurement decisions must be taken without clarity on when conformity assessments and post market obligations will come into effect;
While the Digital Omnibus may ultimately smooth implementation, in the near term it adds a layer of complication to strategic decision making for providers and deployers of AI systems falling within potential high risk use case.
What Should Clients Do Now?
Pending further clarity, organisations should consider the following steps:
- Adopt a principles based compliance approach that focuses on the AI Act’s core requirements—risk management, data governance, human oversight, transparency and accountability.
- Prepare adaptable compliance strategies that can be adjusted quickly when Commission guidelines or harmonised standards are published.
- Identify systems that may fall within Annex III or regulated product categories so that compliance work can be prioritised once timelines become more certain.
- Track Commission publications, standards body activity and progress of the Digital Omnibus through the legislative process.
- Prepare contingency plans for both delayed application and accelerated enforcement following the publication of guidance.
We will continue to monitor developments and provide updates as greater clarity emerges on the revised AI Act implementation timetable.
