First fines under the AI Act
The European Commission has announced the imposition of the first financial penalties linked to the European regulation on artificial intelligence (AI Act), which has been gradually entering into force since 2024. Three companies are being targeted for failing to comply with the transparency obligations imposed by the text.
The fines, ranging from €2 million to €15 million, concern violations related to the absence of technical documentation, failure to label AI-generated content, and insufficient risk assessments for systems classified as high-risk.
What violations exactly?
According to the Commission's official statement, the identified infringements are:
- Lack of transparency: two companies failed to inform users that they were interacting with an AI system, in violation of Article 52
- Insufficient technical documentation: a company deploying a credit scoring system did not provide the documentation required by Annex IV
- Absence of an impact assessment: a facial recognition system used in a commercial context had not undergone any prior conformity assessment
“These first sanctions send a clear signal: the AI Act is not a theoretical text. Companies deploying AI systems in Europe must comply with transparency and security obligations” — Thierry Breton, European Commissioner.
A gradual implementation timetable
The AI Act applies in stages. The strictest prohibitions (social scoring, subliminal manipulation) have been in force since February 2025. Transparency obligations for general-purpose AI systems have been active since August 2025. The full requirements for high-risk systems will apply in full from August 2026.
Industry reactions
The sector's reaction is divided. European startup associations welcome the clarification, while warning of the risk of overregulation that could hinder European players in the face of American and Chinese competition.
Among the major AI providers, Anthropic, OpenAI and Google have all published AI Act compliance guides and state that their models meet the regulation's requirements.
What this changes for French companies
For companies that use or deploy AI systems, the message is clear: compliance is no longer optional. The priority areas of attention are documenting the AI systems used, informing end users, and conducting impact assessments for any system classified as high-risk.
The coming months will be decisive. The creation of the European AI Office, tasked with overseeing the implementation of the regulation, suggests that the intensity of inspections will increase significantly by the end of the year.
Comments· 2 comments
The article feels a little too thin on what the alleged transparency failures actually involved and how the penalties were calculated. Calling these the first AI Act-related sanctions is significant, but the piece could better explain what this means for smaller companies trying to comply.
I agree that more detail would help, but I think the article’s main value is highlighting that the rules may now have real enforcement consequences. The specific cases may matter less than the broader signal that transparency obligations should not be treated as optional.