AI content: EU adopts mandatory labelling Code

Brussels – The European Commission adopted today (10 June) the final version of the Code of Good Practice on the labelling and tagging of content generated by artificial intelligence (AI). This refers to so-called watermarking and, under the rules of the Artificial Intelligence Act (AIA), it will become a requirement for all providers and users of generative AI systems from 2 August, or 2 December if the revision of the AIA is formally approved.

Drafted by six independent experts and over 180 stakeholders, the Code provides practical measures to help companies comply with the transparency obligations enshrined in the AIA. Specifically, the text comprises two sections. The first is dedicated to the obligations of providers of generative AI systems. It sets out how to ensure that audio, images, videos, or text generated or manipulated by artificial intelligence are clearly labelled. The second focuses on the obligations of those who use these systems. It clarifies the labelling rules they must also follow.

Alongside the Code, the Commission is preparing to draft further guidelines to clarify the scope of the legal obligations and address aspects not covered by the text finalised today. The Commission is also inviting all stakeholders to sign it, while making it clear that adherence to the Code remains voluntary.

“Europeans have a right to know whether what they see, hear or read has been made or altered by AI, especially when such content can shape public debate,” the Commissioner for Digital Technologies, Henna Virkkunen, stressed. “This Code of Practice gives AI providers and deployers a clear and practical way to label AI-generated content and deepfakes,” she added.