IA law: EU states give the green light

The 27 countries of the European Union agreed to regulate artificial intelligence based on the proposal negotiated between the Commission and the European Parliament. This regulation was the subject of intense debates to reconcile supervision and innovation.

During a meeting of permanent representatives of the European Council (Coreper), all European Union countries validated the proposal for a regulation on artificial intelligence. Also called IA Law, it was the object of a common position between the European Commission and the European Parliament last December.

France was quite cautious regarding this first project. Minister Jean-Noël Barrot indicated in a message that the text constituted “a step in a project that began four years ago and that requires additional discussions”. The President of the Republic, Emmanuel Macron, was skeptical, “we can regulate much faster than our competitors, but we have to be at the right speed. If we lose leaders because of this, we will have to go back to this.” A position to protect French companies like Mistral AI or German companies like Aleph Alpha. Negotiations to modify the text towards less regulation, so as not to limit innovation, were not successful.

AI systems classified according to risks

But what exactly does this text contain? First, it imposes obligations based on the risk level of AI systems. The document insists on those considered “high risk” that could cause damage to health, safety, fundamental rights, democracy, etc. For these systems, it will be necessary to carry out an impact analysis before placing them on the market. Furthermore, an obligation of transparency and explainability of models is established.

The AI ​​Law also includes prohibited systems, “biometric categorization systems that utilize sensitive characteristics (e.g. political, religious, philosophical opinions, sexual orientation, race).” There is also the untargeted extraction of facial images from the Internet or via video surveillance to create facial recognition databases. Certain exceptions are possible, especially for law enforcement. The text also provides for the creation of a European entity in charge of verifying the correct application of the IA Law and financial sanctions (up to 35 million euros or 7% of global annual turnover or 15 million euros and 3% of global AC) in case of violation of regulations. After this validation by the 27 Member States, the text will return to the co-legislators for a final vote.

Leave a Comment