The European Parliament has adopted its position on a package of amendments to the EU Artificial Intelligence Act (AIA), pushing back key compliance deadlines and introducing a new prohibition on non-consensual AI-generated intimate images. On 26 March 2026, MEPs voted 569 in favour, 45 against, and 23 abstentions to adopt the position on the simplification (“omnibus”) proposal amending the AIA, commonly referred to as Digital Omnibus on AI.
Delayed application dates for high-risk AI
The core rationale is pragmatic: key technical standards may not be finalised in time for the original deadline of 2 August 2026. To provide legal certainty, MEPs propose 2 December 2027 for high-risk AI systems explicitly listed in the regulation — including those involving biometrics, critical infrastructure, education, employment, and law enforcement — and 2 August 2028 for AI systems covered by EU sectoral safety legislation. Providers of AI-generated content subject to watermarking requirements would have until 2 November 2026 to comply.
Relevance for medical device manufacturers
Notably, MEPs argue that obligations under the AIA can be less stringent for products already regulated under sectoral laws, including medical devices. This principle — avoiding double regulation — will be a critical point to monitor during upcoming trilogue negotiations.
Next steps
With Parliament’s mandate confirmed, interinstitutional negotiations with the Council can now begin. The original August 2026 deadline creates a tight timeline — a political agreement will likely be needed before June 2026.
Criticism from bsi
In the meantime, notified body bsi published a white paper on its position regarding the changes. They raised concerns on different aspects, e.g. a limitation of oversight in conformity assessments and unclear provisions in the context of parallel conformity assessment procedures. The white paper can be downloaded here.
Source: bsi





