A legislative proposal numbered 3149, introduced on September 15, 2026, in France’s National Assembly, seeks to establish market authorization for general-purpose AI models and certain high-risk AI systems. The bill was presented by René Pilato, a member of parliament from Charente, and co-signed by 69 members of the La France insoumise group, including its coordinator Manuel Bompard and president Mathilde Panot. If adopted, the proposal would create a new National Artificial Intelligence Authority (ANIA), which would have the power to impose sanctions of up to 7 percent of a company’s global revenue for violations of its provisions. As a parliamentary initiative from an opposition group, the proposal will only be considered if included in the agenda, for example, in a niche of the group. As of September 25, the legislative file mentions only the submission and referral to the Economic Affairs Committee. The proposal would require prior authorization for AI systems, which differs from the European AI Act, which relies on self-assessment by providers and market monitoring. The rationale for the bill presents it as a “complementary” measure to the AI Act and cites the case of Grok, an AI system linked to non-consensual sexual content, as justification for a “protective framework” at the national level. It explicitly excludes employment, copyright, and ecological footprint issues, directing them to future legislation. To offset the administrative burden, the state would impose an additional tax on the revenue generated in France by multinational companies that provide or deploy AI systems. The authorization requirement would apply to general-purpose AI models and high-risk systems listed in Annex III of the European AI Regulation. These are systems made available to the public, either for free or for payment, as part of a commercial activity. Personal systems used outside of commercial contexts would be excluded. The proposal also excludes the exception in Article 6, paragraph 3, of the AI Act, which allows certain systems not to be classified as high-risk if they do not pose a significant threat to health, safety, or fundamental rights. This means that even systems benefiting from the European exception would need a new national authorization. The proposal would require a new authorization after any substantial modification that could affect the compliance of an authorized system or change its risk profile. The specifics of the authorization process would be set by decree in the Council of State, but the text itself does not establish a deadline for processing. The ANIA would consist of eighteen members, including two deputies, two senators, and the president of the CNIL or their representative. The text acknowledges that the European AI Board has exclusive monitoring powers over general-purpose AI models but states that the ANIA would cooperate with this board and national authorities through agreements. The Economic Affairs Committee would need to determine how a national market authorization would interact with European harmonized rules if the proposal is examined.