European Commission publishes guideline on transparency obligations under AI Act
The European Commission on Monday published new guidelines to assist providers and deployers of artificial intelligence (AI) systems in fulfilling transparency obligations under the EU’s AI Act, including examples of exceptions to these requirements. While not specific to healthcare products, companies developing healthcare products with an AI system component may be subject to transparency rules.
The guidelines specify AI providers’ and deployers’ responsibilities to inform users when interacting with AI systems and to ensure that AI-generated or manipulated content is identifiable under Article 50 of the AI Act. The AI Act entered into force on 2 August 2024, and the transparency provisions are set to take effect on 2 August 2026.
“With today's guidelines, the Commission supports the smooth and effective application of the AI Act to make AI systems interacting with people, such as chatbots and AI agents, and AI content more transparent and trustworthy. These guidelines support providers and deployers in meeting their obligations under the AI Act, while helping citizens know when they are interacting with AI,” announced Henna Virkkunen, Executive Vice-President for Tech Sovereignty, Security and Democracy of the Commission.
According to a Commission press release, “transparency obligations will help people recognise when they are interacting with AI or when content has been generated or altered by AI, reducing the risk of deception and manipulation. The guidelines published today clarify which providers and deployers must comply with the transparency obligations for interactive AI systems and the marking and labelling of AI-generated content.”
The press release said the guidelines clarify several key concepts and provide exemptions to the Act and examples. These concepts include directly interactive AI systems, such as chatbots, as well as synthetic content, which encompasses both partially and fully AI-generated text, deepfakes, and AI-generated text on topics of public interest.
The guidelines require providers to design AI systems so that individuals are explicitly informed whenever they interact directly with an AI system, according to the release. They must also include machine-readable markers to facilitate the detection of AI-generated or manipulated content.
Organizations deploying AI systems must inform individuals when they are exposed to the following: emotion recognition and biometric categorization tools, deepfakes, and text publications on matters of public interest that lack human review or editorial control – unless they meet specific exceptions, such as obviousness.
The transparency requirements are separate from the AI Act’s risk classification system, and compliance with Article 50 does not affect other obligations for high-risk systems under Chapter III of the act.
However, the guidelines note that the transparency requirements under Article 50(1) may be excepted if it is obvious that interactions with the system are artificial. For example, certain systems only meant to be used by trained health professionals to support medical diagnoses and suggest potential treatments would be excepted from Article 50(1) for obviousness.
AI system developers may be excepted from transparency requirements under Article 50(2) if the AI system only makes standard editing or minor alterations. For example, the guidelines state that “medical image technical processing or reconstruction within medical devices” such as annotations, boxes, contours, circles, or heatmaps would be excepted from transparency rules.
Guidelines, Press release; Announcement
