On 16 June 2026, the European Parliament adopted its plenary position on the Digital Omnibus on AI - the package of targeted amendments that simplifies and recalibrates the EU AI Act. The vote follows the provisional political agreement reached between Parliament and Council negotiators on 7 May 2026, and advances the file towards formal adoption ahead of the date on which the AI Act's high-risk obligations would otherwise have begun to apply.
The substance of the package has been clear since the May agreement. For standalone high-risk AI systems classified under Annex III, the application of requirements is deferred from 2 August 2026 to 2 December 2027, while high-risk systems embedded in regulated products under Annex I move to 2 August 2028. The package also introduces mechanisms to limit duplicative obligations where sector-specific legislation already imposes AI-related requirements, alongside the prohibition on AI systems designed to generate non-consensual intimate imagery that featured prominently in the political agreement. The Parliament's adoption of its position is a procedural milestone rather than a change of direction - but it is a necessary step before the amendments can be finalised and published.
Running alongside the legislative package is the EDPB and EDPS's continued scrutiny of the broader Digital Omnibus. In their joint opinions, the two bodies supported the objective of simplification and competitiveness while raising pointed concerns - most notably urging co-legislators not to adopt proposed changes to the definition of personal data, which they warned would narrow the concept well beyond the established case law of the Court of Justice. That tension between simplification and the preservation of protection continues to shape the debate as the files progress.
For organisations building or deploying AI systems, the practical message is unchanged by the procedural step: the extended deadlines redistribute the compliance workload rather than reducing it. The additional time should be used to mature governance, documentation, and risk-classification practices, not to defer them.
Acompli perspective: Parliament's adoption of its position moves the AI Act amendments closer to certainty, but the direction has been set since May and the underlying obligations are not going away. The extended high-risk deadlines give organisations runway to strengthen their risk assessments for AI systems and to document classification decisions properly. Meanwhile, the EDPB's resistance to narrowing the definition of personal data is a reminder that the GDPR's foundations are not being dismantled - so accurate data mapping and well-maintained records of processing remain the baseline, whatever the final shape of the Omnibus.