Macron calls for banning AI that harms children and signs a letter with OpenAI and Anthropic: pre-registration comes to models

By Arab News (AFP) · November 10, 2026.

French President Emmanuel Macron said on Monday that artificial intelligence should be banned wherever it poses a risk to children's mental health. "If there is a risk to children from certain uses of AI, then it must be banned," he said in an interview with the newspaper Les Echos, as reported by AFP. He also argued that AI should not be "the Wild West" and that it should be prohibited when it "threatens" the mental health or brains of minors.

The statement comes just ahead of the adoption of a charter at the Paris Peace Forum. According to Macron's office, the document will bring together OpenAI and Anthropic, among others. On Tuesday, an expert committee will hand the president its recommendations on protection from AI, an area less explored than the impact of social media.

The labs would commit to measures such as independent evaluations of generative AI systems brought to market and access to their models' data. Macron says they will accept "the principle of pre-auditing and pre-certification," meaning a review of models before they are put in children's hands. According to the Élysée, the charter is intended to feed into national and European legislative debates.

Macron framed the goal as a framework covering three categories of danger: children, democratic security and major national security issues. He also convened some 50 heads of state and government, international organizations, companies and researchers by videoconference. Participants included the prime ministers of the United Kingdom, Canada, Norway, the Netherlands and Japan, as well as the CEOs of OpenAI, Anthropic, Mistral and Microsoft. "The diagnosis is shared. We must act together," he wrote on X. All of the above are statements by the president and his office; the charter has not yet been adopted.

**Our take.** The most relevant part is not the line about a ban, which is political, but the mechanism: pre-certification before commercialization. It is the shift in logic already seen with drugs and toys, and one that social media never had: prove the harm beforehand, not tally it afterwards. That OpenAI and Anthropic are sitting at the table suggests they would rather shape that rule than have it imposed on them.

There are reasons for caution. A voluntary charter is not a law, and "independent evaluation" only counts if the evaluators have real access to the models and measurable criteria. In general, the usual risk with this kind of regulation is legislating panic rather than demonstrated capability: "a threat to the brain" is a vague bar until it is translated into concrete tests, such as emotional dependence, harmful content or manipulation in companion chatbots. Our position is that governance should be evidence-based, and protecting minors is the area where consensus is easiest and where evidence, still scarce, is most urgent to generate.

In the short term, this adds friction and compliance costs, and may favor those who already have teams for audits. In the long term, AI that reaches children with safeguards is a precondition for it to deliver on its promise in education and health: personalized tutors and care at scale will only be acceptable if families trust them. That trust is built with credible prior controls, not slogans. The real test will come when the charter becomes European legal text and someone publishes the results of the first audit.