2023-12-11 14:40:13
TECH – After more than 35 hours of negotiations, the European Council and Parliament have agreed on legislation to regulate artificial intelligence (AI). The signing, on Friday December 8, 2023, of the AI Act, legislation governing the use and development of this technology, presented in April 2021 and adopted by the European Parliament last June, is a world first. The text, which intends to limit the potential abuses of AI such as the distribution of misleading content and others deepfakes, prohibits its use for rating systems similar to Chinese social credit, biometric identification or surveillance. Market players, who feared excessive or even repressive regulation, criticized the text.
What does the European regulation on artificial intelligence provide? Once the text enters into force, companies will be obliged to assess the risks of their solutions themselves and to label content generated by AI (images, videos, etc.). The AI Act also provides for the classification of systems according to a level of risk, ranging from “minimal” to “unacceptable” if they present “significant harm to health, safety, maintenance of order, education , the fundamental rights of people or the environment”.
Publishers will also have to provide human control over the machine, establish technical documentation and implement a risk management system.
“Historic” for Thierry Breton
The text was adopted on June 14, 2023 by an overwhelming majority of the European Parliament and since then, European digital giants have been watching the start of negotiations within the European Union (EU).
The ban on biometric identification was proposed by the European Commission in 2021 but MEPs opposed it. Such a practice is reminiscent, in the eyes of many, of mass video surveillance and China’s social credit system.
“Historic! The EU becomes the first continent to set clear rules for the use of AI”, welcomed Thierry Breton, European Commissioner at the origin of the project, in a post X. The challenge for Member States was to achieve regulation that was sufficiently restrictive to prevent abuses without nipping in the bud emerging innovation driven by several startups, such as Aleph Alpha in Germany or Mistral AI in France.
Generative AI was also at the heart of discussions, after the appearance of tools like ChatGPT and GPT-4 from the Californian company OpenAi, or even video, sound and image generators. Hundreds of personalities from the tech world like Elon Musk had called for a moratorium on research into AI. In a petition, the signatories raised the danger that this technology represents for humanity, societies and democracies.
MEPs particularly called for more transparency on the algorithms and databases used by these systems. The IA Act intends to ensure the quality of the data and guarantee compliance with copyright legislation. As for the most powerful systems, other constraints are provided for by the text, in the same way as other regulations such as the Digital Services Act (DSA).
“One step”, puts Jean-Noël Barrot into perspective
In France, the French Minister of Digital, Jean-Noël Barrot reacted to the news. In a much less enthusiastic tone than that of Thierry Breton, he describes the signing of the agreement as “a step”. “France will be attentive to ensuring that the capacity for innovation is preserved in Europe. The best protection we can offer our fellow citizens is to have our own European models of artificial intelligence,” he wrote. He indicated that his services “will very carefully analyze the compromise reached and ensure in the coming weeks that the text preserves Europe’s capacity to develop its own artificial intelligence technologies and preserves its strategic autonomy.” If “Europe missed the GAFAM train, we must get a head start on AI”, welcomes Mr. Barrot.
This agreement is not unanimous. The European head of the Computer & Communications Industry Association (CCIA), Daniel Friedlaender, regretted that “speed has prevailed over quality, with potentially disastrous consequences for the European economy”. “Technical work” East “necessary” on crucial details, he believes.
Negotiations within the EU were being tackled by French and German companies. In an open letter, Bitkom and Numeum, two associations representing the interests of more than 5,000 tech companies in both countries, called on lawmakers to maintain a “risk-based approach to ensure that the Union European Union strengthens competitiveness and innovation potential” in order not to be overtaken by the Americans and the Chinese. Before them, it was the leaders of 150 European companies, including French companies such as Airbus, Blablacar, Capgemini and Ubisoft, who signed an open letter in July warning of the consequences of the AI Act.
The European institutions are also equipping themselves with means of surveillance and sanctions with the creation of a European AI office within the European Commission. This body will be able to impose fines of up to 7% of turnover, with a floor of 35 million euros for the most serious offenses.
Europe therefore seems to be taking the lead in the regularization and standardization of artificial intelligence, subject to a frantic race between China and the West.
#European #Council #Parliament #reach #agreement #Act