BRUSSELS: EU member states intend to apply stricter rules for artificial intelligence (AI) providers much later than originally planned.
The regulations for AI systems posing specific risks are not due to come into force until December 2027 – 16 months later than planned, according to a statement from the EU delegations on Friday.
In November, the European Commission had proposed postponing the rules to give AI providers more time to implement the regulations.
Originally, the European AI Office was due to enforce some of the rules starting in August. Before the amendments to the law can come into force, member states and the European Parliament must still reach an agreement.
The EU AI Act sets out obligations for AI developers such as ChatGPT and Gemini, including requirements for how their models are trained.
The European Commission had boasted of having the world’s first legislation of its kind.
Following the scandal over sexualised images generated by Grok on Elon Musk's X platform, the AI Act is also set to be amended: a ban is intended to ensure that AI does not generate sexual or intimate content without the consent of those involved, nor create depictions of child sexual abuse.
X came under fire at the end of 2025 because, initially, all users could ask the AI chatbot Grok to sexualize posted images. Time and again, people instructed the AI to manipulate photos and dress women in bikinis, for example. – dpa
