ChatGPT and Gemini in France may need approval like a medicine

What would a ChatGPT and Gemini approval involve?

The heaviest proposal would make launching, and then keeping, a consumer generative AI system on the market depend on a prior multidisciplinary assessment of its impacts. It targets only services aimed at the general public, ChatGPT and Gemini first among them. Professional tools would escape it. Benyamina openly defends the pharmaceutical comparison and talks of a real "marketing authorisation".

The report also asks that authorities and accredited researchers get access to training, usage and monetisation data, and that France set up a national protocol for AI-related information crises. The authors say they do not want to hobble the French or European economy, and they urge Macron to make decisions. A day earlier, he had said in an interview that AI should be "banned" wherever it poses a risk to children.

Why psychiatrists worry about chatbots

Their first concern is cognitive weakening. "It's attractive, we delegate skills to it. Since we delegate skills, we don't acquire them," Gaillard says. Benyamina calls it a "cognitive debt".

The report also wants sycophancy, the tendency of chatbots to flatter users at the expense of the truth, and anthropomorphism, giving the machine human traits, tackled at design stage. Gaillard describes "a system that is always available and has no other intention than keeping you online". The text adds protections for minors and tougher action against non-consensual sexual content and child abuse material generated by AI.

According to the document, 33 million people in France used generative AI at the start of the year, fifth highest in the world. No decision has been announced yet. The findings were due in May and arrived six months late.