Representatives of the European Union's decision-making institutions agreed on Thursday to amend the regulation governing artificial intelligence (the AI Act) to prohibit systems that generate sexual content without consent.
Under the agreement, the ban would cover AI systems that can:
- produce content depicting sexual abuse of children;
- create fake images or videos of an identifiable person's intimate body parts;
- depict someone engaged in sexual activity without that person's consent.
The prohibition would not be limited to systems explicitly marketed for those purposes: applications that appear harmless but do not implement “reasonable safety measures” to prevent such abuses would also fall under the restriction. The scope includes images, videos and audio.
Deadlines and compliance
Companies would have until 2 December 2026 to comply with the new rules. The same deadline was set for watermarking AI-generated content. That represents a delay compared with the current regulation but is shorter than the timetable originally proposed by the European Commission during the amendment negotiations.
The compromise also introduces simplifications and easements: certain obligations would be streamlined, some deadlines extended for the sake of simplification, and exemptions for small and medium-sized enterprises (SMEs) would be broadened.
Legal steps remaining
The deal still requires formal approval by the European Parliament (EP) and the Council (the Council of Ministers). The report notes that representatives of both institutions normally assess political balances before final adoption, so the subsequent legal ratification of similar agreements is often a formality.
Background
A few months ago, the chatbot operated by X generated content that included paedophilic material and images of women being "stripped" without their consent; that incident prompted an EU procedure and has helped drive the push for stricter regulation.
The aim of the agreement is to prevent misuse of automated technologies that pose serious personal and safety risks, in particular the mass production of non‑consensual sexual content.
