European Union member states on Friday approved unprecedented global legislation to regulate artificial intelligence (AI), bringing tense negotiations to a close, Belgium, which holds the EU Council presidency, announced.

During a meeting on Friday, ambassadors on the 27th “unanimously confirmed” the political agreement concluded last December between member states and members of the European Parliament, according to X (ex-Twitter) Belgium, in unison with several diplomatic sources.

Some countries, including France and Germany, raised concerns and were taken into account before the text was finalized, diplomats said.

“This law on artificial intelligence is a significant step, establishing the first rules on the planet aimed at artificial intelligence to make it safer and respect fundamental EU rights,” the Belgian presidency said.

European Commissioner Thierry Breton, responsible for the file, hailed the “historic, first in the world” ruling.

“The AI ​​law unleashed passions, and rightly so! Today, member states approved the December political agreement, recognizing the ideal balance reached by the negotiators between innovation and security,” he said.

In April 2021, the European Commission presented its draft Law on Artificial Intelligence. But the arrival in late 2022 of ChatGPT from the California startup OpenAI, capable of writing dissertations, poems or translations in seconds, added a new dimension and sparked discussions to accelerate.

This system, as well as those capable of creating sounds or images, revealed to the general public the enormous potential of AI. But there are also certain risks: the spread of fake photos on social networks has warned about the dangers of manipulating opinion.

Paris and Berlin wanted European legislation to protect startups specializing in artificial intelligence.

In Berlin on Tuesday, German Digitalization Minister Volker Wissing was pleased that he was “able to achieve improvements for small and medium-sized businesses, avoid disproportionate requirements and ensure that we remain internationally competitive”.

As for generative artificial intelligence, those that can generate texts, sounds or images on demand will be imposed – according to the adopted text – on everything to ensure the quality of data used to create algorithms and to check compliance with European copyright law right. Developers must also ensure that the sounds, images and texts they create are clearly defined as artificial.

Hard limits will apply only to the strongest systems.

Systems considered “high risk” are, respectively, in sensitive areas such as critical infrastructure, education, human resources, public order, etc. – a number of obligations will be placed on them, such as the obligation to have people drive the car, draw up technical documentation or establish a risk management system.