EU AI Act Enforcement Begins August 2: First Fines Will Target Chatbot Makers Missing Risk Disclosures

The European AI Office confirmed on Wednesday that enforcement of the EU AI Act's general-purpose-AI provisions will begin on August 2 as scheduled. Companies operating general-purpose AI systems in the European Union must publish a model card, a training-data summary, and a copyright-compliance attestation by the deadline, and the first fines will target chatbot makers that have not done so.

What the Rules Require

Under the Act, any provider of a general-purpose AI system with more than 10 million monthly active users in the EU is classified as a «systemic-risk» provider and must publish: (1) a detailed model card describing capabilities, limitations, and evaluation results; (2) a training-data summary disclosing the major data sources used, at a category level rather than a record level; and (3) a copyright-compliance attestation confirming that the provider has implemented an opt-out mechanism for rights-holders under Article 53(1)(c). The documentation must be filed with the AI Office and made publicly available on the provider's website.

「August 2 is a hard date. The grace period is over,」 said Lucilla Sioli, director of the European AI Office, in a press briefing. «The first compliance actions will focus on the systemic-risk providers whose documentation is incomplete or absent.」

Who Is Most Exposed

Based on the AI Office's published guidance and recent market data, the providers most exposed to immediate enforcement action are the operators of consumer chatbots that have not yet published a model card or a training-data summary. Several major US-headquartered providers have indicated they will meet the deadline, but at least three have signaled that they are still working through the copyright-compliance attestation. Maximum fines under the Act are €15 million or 3% of global annual turnover, whichever is higher.

What's Next

The first round of compliance actions is expected within the first 30 days of enforcement. Beyond that, the AI Office has indicated that enforcement priorities for the remainder of 2026 will be the high-risk provisions of the Act — which take effect in August 2027 — and the prohibited-practices provisions, which have been in force since February 2025.