As of August 2, the European Commission has been granted authority to impose fines on suppliers of general-purpose AI models (GPAI) for violations of the Artificial Intelligence Act.

As of today, the use of AI becomes more transparent in the EU.

Here’s how EU law makes sure AI works for you, in full openness ↓ pic.twitter.com/tfeQ3vOq9m

— European Commission (@EU_Commission) August 2, 2026

On the same date, new transparency requirements came into effect across the EU. Chatbots and other interactive AI systems must inform users that they are engaging with artificial intelligence, while deepfakes and other AI-generated or altered content must be labeled accordingly.

Changes Implemented

The obligations for GPAI model suppliers were initially established on August 2, 2025. One year later, the European Commission received the authority to enforce compliance, including the ability to levy fines, as detailed in its guidelines for providers.

Companies are required to document technical information about their AI tools, disclose specific details to downstream providers, and adhere to EU copyright laws. Providers of the most powerful models that pose systemic risks must notify the European AI Office and take steps to assess and mitigate those risks.

New entrants to the GPAI market after August 2, 2025, must already comply with these regulations, while those entering later have until August 2, 2027, to meet the requirements.

Fines Imposed

A separate penalty regime under Article 101 of the AI Act applies to GPAI model suppliers. The European Commission can impose fines of up to €15 million or 3% of global annual turnover for non-compliance, failure to provide required documentation or information, submission of incorrect data, non-compliance with requested measures, or refusal to grant access to models for assessment.

This differs from the general penalty regime under Article 99, which allows for fines of up to €35 million or 7% of global turnover for prohibited practices, such as social scoring and other banned actions outlined in Article 5.

Other violations under Article 99, including various requirements for high-risk systems and transparency mandates under Article 50, can incur fines of up to €15 million or 3% of turnover. Fines for providing incorrect, incomplete, or misleading information to national regulators can reach up to €7.5 million or 1% of turnover.

For small and medium-sized enterprises, the lower threshold under Article 99 applies. Larger companies will be subject to the higher threshold.

Labeling Requirements

The new transparency rules mandate:

  • clear communication to users when interacting with an AI system directly;
  • notification regarding the use of emotion recognition and biometric categorization systems;
  • prominent disclosure of AI utilization;
  • addition of machine-readable labels to AI-generated or modified content for easier detection.

The European Commission has noted that a transitional period until December is in place for labeling content generated by generative AI systems that were on the market before August 2, 2026.

A voluntary Code of Practice is available for GPAI model suppliers, assisting companies in demonstrating compliance with the AI Act, but it does not impose new obligations or exempt them from legal requirements.

This Code consists of three sections: transparency, copyright, and security and protection. The first two sections apply to all GPAI model suppliers, while the third focuses on the most powerful systems with systemic risks.

Signatories include Anthropic, Google, IBM, Microsoft, Mistral AI, OpenAI, and others. xAI has only agreed to the security section, meaning it must demonstrate compliance with transparency and copyright requirements through other means.

High-Risk Systems Delayed

Regulations concerning high-risk AI systems will take effect on December 2, 2027, and for AI integrated into physical products such as medical devices, toys, and elevators, the deadline is set for August 2, 2028.

The timelines were extended following the implementation of the Digital Omnibus in July 2026. According to the European Commission, this extension is intended to allow more time for the preparation of standards and the establishment of competent authorities across EU member states.

The regulator has also launched a tool for filing complaints regarding alleged violations of the AI Act and a separate tool for whistleblowers professionally connected to GPAI model suppliers.

Notably, in June 2026, Anthropic's CEO Dario Amodei urged the U.S. to tighten its regulatory approach to artificial intelligence, advocating for a shift from mere disclosure to mandatory safety checks for the most powerful AI models.