The European Parliament has approved changes to the EU’s AI Act that delay some compliance deadlines and introduce a ban on so-called “nudifier” apps, following a plenary vote on Monday.
MEPs backed the amendments by 423 votes to 57, with 174 abstentions, the parliamentary press service reported.
Under the new timetable, obligations on “high-risk” AI systems will apply from 2 December 2027 for stand-alone high-risk systems and from 2 August 2028 for AI systems embedded as safety components that fall under existing EU product safety and market surveillance laws.
High-risk systems are those subject to stricter controls under the AI Act because they can affect safety or fundamental rights.
Watermarking rules for AI-generated content — requiring machine-readable labelling — will also be delayed until 2 December 2026.
Ban on “nudifier” tools
The amendments also ban AI systems that generate child sexual abuse material, and tools that create images, video or audio depicting an identifiable person’s intimate parts or sexually explicit activities without their consent, the Parliament said.
Providers will be barred from placing such systems on the EU market unless they include “adequate technical safeguards” to prevent the creation of that material, it added, and the ban also covers users deploying the tools for that purpose.
Companies will have until 2 December 2026 to bring relevant systems into line.
The legislation also removes overlapping requirements for AI used in machinery products by clarifying that they must comply with sector-specific safety rules, and it introduces a clearer definition of what counts as a “safety component”.
The package includes an option to process personal data where “strictly necessary” to detect and correct bias in AI systems, with safeguards, and extends some exemptions previously available to small and medium-sized enterprises to small mid-cap companies.
Enforcement for certain general-purpose AI systems will also be streamlined within the EU’s AI Office.
Co-rapporteur Arba Kokalari said Parliament was “pressing the pause button on the AI Act” by extending timelines and reducing duplication of rules for machinery products, in the debate published by the European Parliament.
Co-rapporteur Michael McNamara said the changes included “an outright ban on AI nudification apps”, adding that the ban would enter into force before the end of the year.
The amended law still needs to be formally adopted by the Council before it can enter into force, the European Parliament said, while most provisions of the original AI Act are due to start applying on 2 August 2026.
