Guidelines on the implementation of the transparency obligations for certain AI systems under Article 50 of the AI Act — C(2026) 5054 final, Annex

(26) Article 50 AI Act applies to certain AI systems, which includes general-purpose AI systems (‘GPAI systems’) (15), such as chatbots supporting direct interaction with natural persons or AI systems capable of synthetic content generation or manipulation (including deep fakes) (16).

(27) Article 50 AI Act does not explicitly apply to GPAI models. However, if an interactive or generative AI system that falls within scope of Article 50(1) or (2) AI Act is built upon a GPAI model provided by the same provider, the transparency measures set out in these Guidelines could also be implemented at the model level (see also Section 4 below). Furthermore, while they fall outside the scope of Article 50 AI Act, providers of other GPAI models and of other generative AI models are also encouraged to implement appropriate transparency measures at the model level for the identification of interactive AI systems and the marking of AI-generated and manipulated content. This is particularly beneficial to facilitate compliance by downstream AI system providers with their obligations under Article 50(1) and (2) AI Act (17) and to contribute to preserving the trust and integrity of the information ecosystem. For GPAI models presenting systemic risk, model-level safety measures for the identification of AI interactions and marking of AI-generated or manipulated content may, where appropriate, constitute part of a broader set of mitigating measures that providers may take as part of their obligation to mitigate systemic risks pursuant to Article 55(1)(b) AI Act.