Digital Omnibus

High-risk AI systems: requirements postponed until 2027

Among the amendments approved by the European Parliament are a ban on regulatory overlaps for machinery and a ban on nudification apps

by Camilla Colombo

17 June 2026

epa12918351 Una veduta dell'emiciclo del Parlamento europeo a Strasburgo, in Francia, il 28 aprile 2026. L'attuale sessione plenaria si svolge dal 27 al 30 aprile 2026. EPA/RONALD WITTEK EPA

3' min read

Translated by AI

Versione italiana

Key points

The postponement
Ban on ‘nudity’
Reducing regulatory and bureaucratic burdens
Reactions

3' min read

Translated by AI

Versione italiana

The European Union’s drive towards simplification of the rules and regulations relating to artificial intelligence continues. The legislation, which was finally approved yesterday by the European Parliament, forms part of the seventh simplification package proposed by the European Commission on 19 November 2025, the so-called Digital Omnibus on AI (Omnibus VII), to assist businesses in implementing the Artificial Intelligence Act (2024/16899). In particular, the amendments concern the deferral of certain obligations for high-risk AI systems; the removal of regulatory overlaps concerning machine safety; and the ban on tools for “nudification” and the creation of child sexual material generated by artificial intelligence.

The postponement

The measures to defer and reduce obligations for systems using AI, adopted with 423 votes in favour, 57 against and 174 abstentions, ease the red tape imposed on businesses and extend the implementation deadlines, to ensure that the necessary standards and support measures are put in place first, whilst retaining the key provisions and the risk-based approach that have characterised the AI Act.

The requirements for high-risk systems will apply:

from 2 December 2027 for high-risk standalone AI systems;
from 2 August 2028 for AI systems integrated as safety components and governed by EU sector-specific legislation on safety and market surveillance.

The law also postpones the implementation of marking and labelling requirements for AI-generated content, in a clear and legible manner, to increase transparency.

Ban on ‘nudification’

Also by 2 December next, businesses must comply with the law that prohibits AI systems that generate child sexual abuse material or create images, videos and audio depicting the private parts of an identifiable person or sexually explicit acts without their consent. Suppliers will not be permitted to place these systems on the EU market unless they are equipped with adequate technical safeguards to prevent such content from being created. The ban also applies to users.

Reducing regulatory and bureaucratic burdens

The other amendments to the AI Act begin with the removal of regulatory overlaps for machines that use artificial intelligence, clarifying that manufacturers will only have to comply with sector-specific safety regulations, and extend to small and medium-sized enterprises (SMC) of the exemptions provided for SMEs under certain regulations, in order to support their growth. In addition, there is a revision of the definition of “safety component”, meaning that products with AI functions that are limited to assisting users or optimising performance will not automatically be subject to the obligations applicable to high-risk systems, provided that their failure or malfunction does not pose risks to health or safety. There is also the possibility of processing personal data where strictly necessary to identify and correct distortions, with appropriate safeguards in place for both high-risk and non-high-risk systems. Finally, the application of the rules for certain general-purpose AI systems will be simplified through the European Office for Artificial Intelligence.

Before it can come into force, the Act must be formally adopted by the Council. Most of the provisions of the Ai Act, which have not been amended during this legislative process, will come into force on 2 August.

Reactions

During the plenary debate, the co-rapporteur for the Committee on the Internal Market and Consumer Protection, Arba Kokalari, was very clear about the reasons that led Parliament to adopt these amendments. ‘With these measures, we are putting the implementation of the AI Act on hold and reducing the administrative burden on entrepreneurs and engineers. It must be easier to set up the tech companies of the future and enable them to remain in Europe. We are also eliminating duplication of regulations for AI-based machines and reducing reporting obligations to help businesses.”

The co-rapporteur for the Committee on Civil Liberties, Justice and Home Affairs, Michael McNamara, focused instead on the ban on ‘nudification’ apps: ‘These apps target real people – the vast majority of whom are women – with the aim of humiliating, degrading and objectifying them. I am proud that this Parliament has fought for this ban, which will come into force before the end of this year.’