Brussels (dpa) – In anticipation of the upcoming EU regulations for large AI models such as ChatGPT and Gemini, the EU Commission has released guidelines aimed at assisting general-purpose AI model providers in meeting transparency requirements outlined in the EU AI Act. The guidelines detail which models and providers are subject to these new regulations and the documentation they are required to maintain.
Starting August 2, 2025, specific criteria will be enforced across the EU for these versatile systems capable of producing text, generating language, or creating images and videos. Providers must disclose operational mechanisms, the data used for training, and their approach to managing potential risks.
Phased introduction of regulatory oversight
Models developed with considerable computational resources will be included under these regulations. For particularly advanced models that pose systemic risks to fundamental rights or security, there will be additional requirements for risk assessment and mitigation.
The Commission emphasizes its goal of providing legal clarity through these guidelines to foster innovation. While the obligations will take effect in August 2025, regulatory oversight will be implemented incrementally: in 2026, the newly established EU AI Office will begin reviewing new models, and in 2027, it will also review existing models. Noncompliance could lead to fines. Meanwhile, the Commission intends to collaborate with providers to ease the transition until the regulatory framework is fully operational, according to EU officials. (July 18)
Leave a Reply