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

(75) Providers of AI systems falling within scope of Article 50(2) AI Act must ensure that the output of their systems are detectable as AI-generated or manipulated content. This means that the provider is obliged to ensure that the means of detection are available to the persons potentially exposed to the content. Furthermore, pursuant to Article 50(5) AI Act, such detection solutions should be able to produce human-readable results indicating whether the content has been AI-generated or manipulated.

(76) A detection tool is a mechanism that detects whether content has been AI-generated or manipulated, typically identifying technical markers or signatures that verify its origin. To ensure full interoperability, providers must rely on publicly-available industry standard detection solutions that allow any third party to implement detection that is ideally locally executable on the digital device and allow seamless integration into self- contained, isolated networks. Where such standards are not available, in particular at the initial stage of the implementation of Article 50(2) AI Act for watermarking technologies, the provider may rely on its own detection solution or on a third party or shared detection solution, so long as those solutions ensure interoperability with the detection solutions used by other providers (see Section 4.2.3.). That possibility should be limited in time until harmonised standards and a standardised provider-agnostic interoperable detection solution emerge that is secure, privacy-preserving and locally executable.

(77) Article 50(5) AI Act further specifies that the information referred to under Article 50(2) AI Act should be provided to the natural persons exposed to the content in a clear and distinguishable manner at the latest at the time of the first interaction or exposure. This information is notably the result of the detection, which indicates whether content is AI- generated or manipulated and not the AI-generated or manipulated content itself. ‘At the time of first exposure’ should be understood as the moment at which natural persons are willing to verify the origin of a specific piece of content and access the detection solution to obtain the detection results.

(78) If the provider of the AI system depends on another actor for the provision or access to a detection solution, it must ensure that the detection solution complies with Article 50(2) and (5) AI Act and that those actors display the detection results in compliance with Article 50(5) AI Act (see Section 7).