On Thursday 26 March, the European Parliament approved an EU ban on artificial intelligence services that allow people to be depicted naked without their consent, confirming – by 569 votes to 45 with 23 abstentions – the position already advocated by the parliamentary committee (see EUROPE 13831/10).
According to a press release, the aim is to ban ‘nudification’ systems, which use AI to create or manipulate sexually explicit or intimate images that resemble an identifiable real person, without that person’s consent. AIs with effective safety measures in place to prevent this technique are not affected.
These initiatives follow on from the introduction of a feature in Grok, the artificial intelligence assistant on the X network, which enabled users to ask it to create hyper-realistic montages (deepfakes) of nude women and children, based on real photos.
In addition, the adopted proposal will delay the application of certain rules relating to high-risk artificial intelligence systems to ensure that guidance and standards to help businesses implement them are ready. For the high-risk AI systems specifically listed in the regulation (including those involving biometrics and those used in critical infrastructure, education, employment, essential services, law enforcement, justice and border management), MEPs propose a date of 2 December 2027.
For AI systems covered by European sectoral legislation on safety and market surveillance, MEPs propose a date of 2 August 2028.
MEPs are also in favour of giving suppliers until 2 November 2026 to comply with the rules on digitally marking audio, image, video or text content created by AI to indicate its origin.
MEPs are also in favour of allowing service providers to process personal data in order to detect and correct biases in AI systems, but have introduced safeguards to ensure that this is only done when strictly necessary.
To help European companies grow as they move beyond the status of small and medium-sized enterprises (SMEs), for which they benefit from certain support measures, MEPs backed the proposal to extend these measures to small mid-capitalisation companies.
Negotiations with the Council of the EU got underway immediately afterwards, with an inaugural trilogue enabling the co-legislators to present their respective positions and mandate the teams of experts to begin the technical work.
The EU Council adopted its negotiating mandate on the AI proposal on 13 March, broadly retaining the essential elements of the Commission’s proposal. It has also included, in Article 5 of the AI Act, a new prohibition on AI practices relating to the generation of non-consensual sexual and intimate content or child sexual abuse material.
A second political trilogue is scheduled for 28 April.
Link to the report: https://aeur.eu/f/lcx (Original version in French by Solenn Paulic)