The European Commission has begun enforcing fines for non-compliance with AI regulations, enhancing transparency in AI usage across the EU.
Effective August 2, the European Commission has the authority to impose fines on developers of general-purpose AI models (GPAI) who do not comply with 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 ↓
— European Commission (@EU_Commission) August 2, 2026
This day also marked the introduction of new transparency requirements, obligating interactive AI systems, including chatbots, to inform users they are engaging with AI technologies, and necessitating clear labeling for deepfakes and other altered content.
Implementation Details
The obligations for GPAI developers were set to commence on August 2, 2025, with the European Commission gaining enforcement capabilities one year later. Companies must maintain comprehensive documentation regarding their AI systems, communicate essential details to downstream providers, and comply with EU copyright laws. Those with models that may pose systemic risks are required to alert the European AI Office and take preventive measures.
New GPAI market entrants from August 2, 2025, are expected to comply immediately, while later arrivals have until August 2, 2027, to align with the regulations.
Penalty Structure
The European Commission can levy fines under Article 101 of the AI Act, with penalties reaching up to €15 million or 3% of global annual revenue for various compliance failures, including incorrect documentation or denial of access to models. This penalty system differs from Article 99, which imposes fines up to €35 million or 7% of global revenue for prohibited practices, with other violations under this article attracting fines similar to Article 101.
Small and medium enterprises will face a more lenient threshold, while larger firms are subject to higher fines.
Transparency Labeling Obligations
New rules demand:
- clear user notifications during AI interactions;
- information on emotion recognition and biometric systems;
- visible disclosure of AI involvement;
- machine-readable labels for AI-generated or altered content.
A transitional period extending until December aims to facilitate compliance for generative AI products launched before the current regulations.
A voluntary Code of Practice is available for GPAI suppliers to assist in compliance, covering sections on transparency, copyright, and security, although participation does not exempt them from mandatory laws. Signatories of this code include major companies like Google, Microsoft, and OpenAI.
Future Regulations and Delays
Regulations for high-risk AI systems are now set to take effect on December 2, 2027, with rules for AI integrated into physical products due by August 2, 2028. These delays follow the Digital Omnibus implementation in July, allowing more time for standard preparations and authority establishment across EU member states.
The European Commission has also introduced a complaint tool for reporting AI Act violations and a whistleblower tool for those associated with GPAI suppliers. In related discussions, Anthropic's CEO Dario Amodei has urged for stricter AI regulations in the U.S., calling for mandatory safety checks for advanced AI models.
