MEPs have adopted rules to regulate artificial intelligence (AI) systems such as ChatGPT - the only legislation of its kind in the world, AFP reports.
Thierry Breton, the European commissioner responsible for the text, welcomed in X the "massive support" of the Parliament (523 votes for, 46 against) for this text, which establishes "the world's first binding and comprehensive rules for trustworthy AI."
The draft law was presented by the European Commission in April 2021. The emergence at the end of 2022 of ChatGPT from the Californian start-up OpenAI, capable of writing essays, poems or translations in a matter of seconds, gave it a new dimension.
This system revealed the enormous potential of artificial intelligence, but also its risks.
The proliferation of fake pictures and videos has highlighted the danger of manipulating public opinion.
"Today is a historic day in our long journey towards the regulation of AI," the co-rapporteur of the text, Brando Benifei (Social Democrats), told the media.
With this text, "we have managed to strike a very fine balance between the interest in innovation and the interest in protection", added co-rapporteur Dragos Tudorace (Renovators, Centrists and Liberals).
However, this legislation "is just the beginning", he said, stressing that artificial intelligence continues to develop rapidly.
"In the future, we will have to pay close attention to this technological development and be ready to respond to any new challenges that may arise," Tudorache warned.
The legislation provides for a two-tier system. AI models for "general use" will have to comply with transparency obligations and European copyright rules.
Systems that are considered "high-risk" - used in critical infrastructures, education, human resources and law enforcement, for example - will be subject to stricter requirements.
For example, they will have to include a mandatory fundamental rights impact assessment.
Artificially generated images, texts or videos (deep fakes) will have to be clearly identified as such.
The text also bans the citizen assessment and mass surveillance systems used in China, as well as the remote biometric identification of people in public places.
However, on the latter point, member states have been granted exemptions for certain law enforcement missions, such as preventing a terrorist threat or targeting victims.
European legislation will be provided with the means to monitor and enforce sanctions with the creation of a European AI Office within the European Commission. It will be able to impose fines ranging from €7.5 million to €35 million, depending on the offence and the size of the company.
"We regulate as little as possible, but as much as necessary," wrote European Commissioner Thierry Breton in X. / BGNES
