Despite criticism, France maintains pressure on AI Act

The AI ​​Act is moving forward, step by step, but one thing does not change: France criticizes the proposed European regulation on Artificial Intelligence (AI). “There are still some issues that we find problematic”, We explain the office of the Minister of Economy, Bruno Le Maire. This is the message that French representatives brought to the meeting of the Council of twenty-seven member states on Wednesday, January 24.

Also read: Articles are reserved for our subscribers Artificial Intelligence: France has not given up on relaxing the AI ​​Act

However the legislative process of the AI ​​Act is almost complete: on Friday 2 February, countries will, in principle, have to decide on the ratification of the text. Despite this, Paris keeps up the pressure: “France’s position not yet decided”

, We assure Bersi, doubts are hanging over the vote which is a cause of serious concern in the cultural industries.

Since June 2023, France has reiterated that some provisions of the text may hinder innovation and European AI start-ups, such as the French Mistral AI, LightOne or Hugging Face, to compete with the Americans OpenAI or Google. Despite a political agreement between the Parliament, the Council and the Commission sealed in Brussels on December 8, 2023, Paris maintained its criticism.

The last stand?

“Regulating foundation models over other countries is not a good idea”, President Emmanuel Macron then explained by referring to the large software programs capable of generating text or images, on which services such as the ChatGPT chatbot depend. Bursey has since tried to influence technical meetings to calibrate the text. But this ended on Friday 19 January. And Paris remains on the offensive.

Also Read | Articles are reserved for our subscribers Artificial intelligence: “France is far from alone in thinking that this is an issue of sovereignty”

What are the issues that still pose problems for France? “On copyright, which is an essential topic, we must find other ways to enforce the production secrets of AI models without making them public.” Supports Mr. Le Maire’s cabinet. According to the text, manufacturers have to publish that “Sufficiently Detailed Summary” Data used to train their software. This should allow rights holders to check whether their texts or images appear there, remove them or negotiate remuneration.

However, after companies like Mistral AI complained about their disclosure “Crafting Recipes” For their competitors, a mention of respect “Trade Secret” was presented. “This is going in the right direction, According to the Ministry of Economy, which worked for this development. But we would like it to change a bit more, for example on the conditions of application.” France wants data summaries to be shared only with a “Trusted Third Party”, like the future “European AI Office”. This will inform the rights holders on request. She had already mentioned this idea at the end of December 2023, without winning her case.

You have 50% of this article left to read. The rest is reserved for subscribers.