Macron says AI should be banned wherever it puts children at risk, and OpenAI and Anthropic are due to sign on to pre-audits
The French president told Les Echos that risky AI uses for kids must be prohibited. A charter naming OpenAI and Anthropic is due at the Paris Peace Forum on November 10. Most of its detail isn't public yet.
French President Emmanuel Macron said on Monday, November 9, that AI should be banned wherever it threatens the mental health of children. “If there is a risk for children with certain uses of AI, then it must be banned,” he told the French daily Les Echos, as quoted by Le Monde.
He added a line that will travel well: “Just because it’s AI doesn’t mean it should be the Wild West.”
What’s actually on the table
The interview was timed to a charter due to be adopted at the Paris Peace Forum on Tuesday. Macron’s office says the text is meant to inform national and European lawmaking, and that it will bring in OpenAI and Anthropic among others, according to the AFP report carried by Arab News. The other signatories haven’t been named.
Macron described what the companies would accept: “the principle of a pre-audit and pre-certification, with a prior review of the models before putting them in children’s hands.” Le Monde adds that the labs would also commit to independent evaluations of generative AI systems on the market and to giving access to their models’ data.
Note the wording. Macron says the companies will accept the principle; the charter itself hasn’t been published, and every account so far comes from his interview and his office. That’s a commitment to a principle. No one has published the wording, a deadline or an enforcement mechanism, so I can’t tell you what a “pre-certification” would involve or who would run it.
The reporting so far, laid out by claim and by who reported it:
| Item | What’s reported | Source |
|---|---|---|
| Ban threshold | Risk to children’s mental health or brain | Le Monde, Les Echos quotes |
| Company commitment | Pre-audit and pre-certification in principle | Le Figaro |
| Named participants | OpenAI and Anthropic | Macron’s office, via AFP |
| Expert committee | Due to give recommendations Tuesday | Le Monde |
| Wider call | Monday videoconference with the UK, Canada, Norway, Netherlands, Japan and the CEOs of OpenAI, Anthropic, Mistral and Microsoft | AFP via Arab News |
Macron also said the framework should cover three kinds of danger: children, democratic security and major national security issues. He called it “the beginnings of frontier AI regulation” and said France would push Europe, and then the Americans and Canadians, to follow.
Why a US reader should care
My read: a prior-review rule for models aimed at minors is a different animal from what Washington has so far. The White House accord signed in September was voluntary and about 300 words long, which I covered in Trump’s AI accord is about 300 words long. Eight US states are also in court with OpenAI over kids and ChatGPT, as in Eight states sued OpenAI over kids and ChatGPT. My read: a charter that two US labs sign in Paris gives European lawmakers a document to point at, even if it binds nobody yet.
My read: the signature is the easy part. Companies agree to “principles” at forums all the time, and the fight starts when someone has to define the risk, run the audit and decide what happens to a product that fails it. I’d wait for the actual text before calling this regulation. If the charter does include access to model data for outside researchers, that would be the part with real teeth, and it’s the part I’d check first.
What to watch: the expert committee’s recommendations and the charter text, both due Tuesday.