THE WORLD IS NOT STANDING STILL.RSS
BIG CHANGE.
BIG CHANGE, one tap away
BIG CHANGE, one tap away

Make room for the big picture.

Add BIG CHANGE to your home screen or desktop. It opens in its own window, with the latest stories when you’re online.

iPhone & iPad
Open bigchange.ai in Safari. Tap Share, then Add to Home Screen. Turn on Open as Web App if shown, then tap Add.
Android
Open bigchange.ai in Chrome. In the browser menu, choose Install app or Add to Home screen.
Desktop
In Chrome or Edge, use the install icon in the address bar or the browser’s app menu. In Safari on Mac, choose File → Add to Dock.

Markdown edition

# French experts propose safeguards for generative AI

> A commission sets out child protection and evaluation proposals.

By BIGCHANGE.AI Editorial

Published: 2026-11-10T07:28:38.104Z
Updated: 2026-11-10T07:28:38.104Z
Canonical: https://bigchange.ai/blog/france-generative-ai-vulnerabilities-report

![An adult and child look at a tablet together on a sofa, beside a soft toy.](https://bigchange.ai/api/media/file/family-tablet-concept.png)
AI-generated illustration for BIG CHANGE

A French expert commission has proposed 18 measures on generative AI, including child safeguards and pre-market checks for consumer chatbots. The [report](https://www.elysee.fr/admin/upload/default/0001/20/cb2da2d0eb9e830ef129b49cde31497efd577271.pdf) was presented on November 10.

The commission’s recommendations are advice, not enacted rules.

## What the commission proposes

The report calls for pre-market assessment of consumer conversational AI, followed by continuing checks after launch. It proposes independent, multidisciplinary evaluation and controlled access to usage, training and monetisation data for authorised researchers and authorities. Assessments would examine effects on vulnerabilities before release and after substantial changes.

For services accessible to minors, it recommends safeguards including filtering dangerous requests. It also proposes prior authorisation for toys incorporating generative AI and a consensus process to determine age thresholds.

The [EU AI Act](https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai) already prohibits certain harmful manipulation and exploitation of vulnerabilities. Those prohibitions have defined conditions; they do not mean every flattering chatbot response is unlawful.

## A question of evidence and privacy

A [survey](https://files.interface-eu.org/files_for_publications/2026/261022-teens-generative-ai-child-safety-Jessica/YouGov-Interface_EnfantsGenAI_FR_FINAL.pdf) by YouGov for Interface offers one view of use, not a clinical assessment. It surveyed 970 French adolescents aged 11 to 17 online from June 24 to 30, with parental consent, alongside 1,138 parents.

Among the 764 young users in the question’s base, 92% said they spent more time than desired on at least one of the tools surveyed. The survey grouped general-purpose chatbots, companions, social-media AI and voice assistants together. The finding is self-reported, applies to that user subgroup and does not diagnose addiction or establish that AI caused harm.

The same study found that 89% of adolescents familiar with AI identified at least one positive aspect. Its parent sample was constructed for demographic representativeness; it does not make the adolescent responses a clinical or causal study.

In separate [guidance on AI and children](https://www.unicef.org/innocenti/reports/policy-guidance-ai-children), UNICEF calls for children’s safety, privacy and preparation for AI. It describes both opportunities and risks, rather than treating every use as harmful.

Designing age checks also involves a privacy trade-off. France’s data-protection authority [CNIL](https://www.cnil.fr/fr/verification-de-lage-en-ligne-trouver-lequilibre-entre-protection-des-mineurs-et-respect-de-la-vie) warns that verification can change privacy protections when access otherwise requires no identification. In its guidance on restricted online content, CNIL describes separating the issue of an age credential from its transmission to a site. An independent intermediary can help limit the identity data sent to the service. That is a privacy design option for age checks, not an implementation chosen for these proposals. A workable policy would still need to define what is verified, by whom and what information is retained.

Any practical effect depends on whether and how policymakers adopt these recommendations.

## Sources

- [Commission report](https://www.elysee.fr/admin/upload/default/0001/20/cb2da2d0eb9e830ef129b49cde31497efd577271.pdf) — Recommendations and cited evidence.
- [French presidency announcement](https://www.elysee.fr/emmanuel-macron/2026/11/10/remise-du-rapport-de-la-commission-dexperts-ia-generative-et-vulnerabilites) — Release and report context.
- [UNICEF guidance on AI and children](https://www.unicef.org/innocenti/reports/policy-guidance-ai-children) — Children’s safety and privacy context.
- [CNIL age verification guidance](https://www.cnil.fr/fr/verification-de-lage-en-ligne-trouver-lequilibre-entre-protection-des-mineurs-et-respect-de-la-vie) — Privacy and verification trade-offs.
- [YouGov survey for Interface](https://files.interface-eu.org/files_for_publications/2026/261022-teens-generative-ai-child-safety-Jessica/YouGov-Interface_EnfantsGenAI_FR_FINAL.pdf) — Method, question and reported usage.
- [European Commission AI Act overview](https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai) — Existing risk-based rules and prohibited practices.