Starting August 2, the European Commission has officially been granted the authority to apply financial sanctions to providers of general-purpose AI (GPAI) models for violations of the Artificial Intelligence Act (AI Act). This is a landmark moment for the entire industry: the regulator is moving from declarative requirements to actual enforcement.
As of the same date, new transparency rules come into force. Chatbots and other interactive AI systems are required to inform users that they are interacting with artificial intelligence. Deepfakes and any content created or altered by AI must receive appropriate labeling.
Key changes for GPAI providers
Obligations for providers of GPAI models have been in effect since August 2025, but now the European Commission has the tools for their enforcement. Companies are required to document the technical characteristics of their AI tools, disclose data to downstream providers, and comply with EU copyright rules. For the most powerful models with systemic risks, notification to the European AI Office and the implementation of measures to assess and mitigate these risks are required.
Fine amounts: what violators face
A separate regime applies to GPAI providers under Article 101 of the AI Act. Violating obligations, refusing to provide documents or access to the model for assessment can result in a fine of up to €15 million or 3% of global annual turnover. This is fundamentally different from the general regime under Article 99, where prohibited practices—such as social scoring or manipulative systems—carry fines of up to €35 million or 7% of turnover.
It is also worth noting that providing incorrect or misleading information to national regulators carries a fine of up to €7.5 million or 1% of turnover. For small and medium-sized businesses, the lower of the two thresholds applies, providing some relief for startups.
Labeling requirements and transition periods
The new rules require:
- clearly informing users about interaction with an AI system;
- informing about the use of emotion recognition and biometric categorization systems;
- disclosing the fact of AI usage;
- adding machine-readable labels to content created or altered by AI.
For content created by generative AI systems placed on the market before August 2, 2026, a transition period applies until December. This gives developers time to adapt without rushing.
Code of Practice and industry positions
The European Commission is actively promoting a voluntary Code of Practice for GPAI providers. It has been signed by Anthropic, Google, IBM, Microsoft, Mistral AI, and OpenAI. However, xAI has only agreed to the safety chapter, leaving transparency and copyright issues to be addressed through independent demonstration of compliance—this creates an interesting precedent for the market.
High-risk AI systems will only begin to be regulated in December 2027, and AI in physical products—from August 2028. The European Commission has also launched tools for filing complaints and whistleblowing, strengthening oversight.
My comment: This step by the EU is a signal to the entire industry: the era of self-regulation is coming to an end. Companies, especially those working with GPAI models, should already be reviewing their compliance procedures. Fines of 3% of turnover are not a formality but a serious financial risk, especially for major players. The market is awaiting the first wave of real AI Act enforcement cases, and their outcomes will set the tone for years to come.