EU member states and Parliamentarians of the European Parliament have reached a landmark agreement on artificial intelligence (AI) rules as part of the European Commission’s so-called digital omnibus package.
The European Parliament said in a press release Thursday morning after late-night talks that the temporary agreement includes delaying the implementation of some obligations for AI systems to prevent legal uncertainty.
What are the amendments to the EU Artificial Intelligence Act?
An agreement was reached to postpone rules on biometrics or high-risk AI systems related to critical infrastructure, education, employment, law enforcement and border management to December 2, 2027, from the earlier deadline of August 2 this year.
It was also agreed to exclude machinery from the AI Act, with the idea that it would only need to comply with the regional safety rules to which it is already subject.
The move by EU co-legislators is being seen by critics as bowing to pressure from businesses.
Mandatory watermarking of AI-generated content will come into effect from December 2.
“Today’s agreement on the AI Act significantly supports our companies by reducing recurring administrative costs. It ensures legal certainty and a smoother and more harmonious implementation of the rules across the Union, strengthening the EU’s digital sovereignty and overall competitiveness,” Cyprus Deputy Minister for European Affairs Marilena Rouna said in a statement.
Cyprus currently holds the rotating presidency of the Council of the European Union.
Ban on “nudification” apps
EU member states and European Parliament negotiators also agreed to ban AI applications used to create unauthorized sexually explicit deepfakes.
Content may include images, video or audio. Companies will have until December 2 this year to bring their systems back on line.
The planned ban also aims to cover the creation of material that clearly depicts child sexual exploitation.
With its move, the EU is, among other things, responding to incidents earlier this year, when users used US billionaire Elon Musk’s chatbot Grok to generate and disseminate millions of sexually explicit AI-generated images of women and children online.
“We are enhancing the protections we provide to children by targeting the risks associated with AI systems. This agreement is clear evidence of our institutions’ ability to act quickly and meet our commitments,” the statement said.
Simplification measures
The EU’s AI Act comes into force in August 2024, with key provisions to be implemented in gradual phases.
Wednesday’s amendments are part of the European Commission’s drive to simplify new digital rules.
The agreement will now require formal approval by a plenary session of the European Parliament and from EU governments, which is generally considered a formality.
Edited by: Dmytro Lyubenko