EU eases AI regulation: Key rules delayed

The changes are part of a broader initiative by the European Commission to simplify new digital regulations, following complaints by European companies that overlapping rules and administrative burdens put them at a disadvantage compared to American and Asian competitors.

According to the agreement reached, the application of the rules for high-risk artificial intelligence systems — including biometric technologies and systems related to critical infrastructure and law enforcement — has been postponed until December 2, 2027. The previous deadline was August 2 of this year.

Exemption of machines from the application of the Law on Artificial Intelligence was also agreed because they are more covered by sectoral regulations, the EU has partially met the requirements of the industry.

At the same time, the negotiators supported the ban on AI systems that generate unauthorized sexually explicit content and the so-called deepfake, i.e. artificially generated content in which the person from the original recording or photo is replaced with another person due to the spread of false information or deception. The decision comes after controversies related to the content created by Elon Musk’s xAI chatbot Grok on the X platform. The ban will apply from December 2.

“By the end of this year, everyone, especially women and girls, will be safer than the terrible apps for generating nude images that are widely available on the EU market,” said Dutch MEP Kim van Sparrentak.

From December 2nd, mandatory watermarking of content generated by artificial intelligence will be applied.

Despiteč relaxation of certain provisions, the European Law on Artificial Intelligence is still considered the strictest regulatory framework for AI in the world.