EU AI Act Article 50(II)

Article 50(II) of the EU AI Act is the transparency provision requiring disclosure of AI-generated or AI-assisted content. The CEAVERS site labels AI provenance on every page and documents the policy in full at /editorial-policy/.

The legal obligation

Article 50 of the EU AI Act addresses transparency obligations for certain AI systems. Paragraph II specifically requires providers and deployers of AI systems that generate or manipulate content — including text, images, audio, and video — to ensure that the output is marked as artificially generated or manipulated in a machine-readable format, and disclosed to natural persons in a way that is clear and distinguishable.

The obligation applies to systems that generate synthetic audio, video, image, and text content intended to resemble real persons or events. For text-generating AI systems used in publishing contexts, the practical obligation is disclosure: readers must be told when content has been AI-generated or substantially AI-assisted.

When it applies

The transparency provisions of the EU AI Act entered force progressively from 2024, with Article 50 applying from August 2026. The General-Purpose AI Code of Practice, which governs the largest AI providers, was finalised in mid-2026. For publishers and research organisations using AI tools in content production, August 2026 is the compliance deadline for implementing disclosure mechanisms.

What disclosure requires in practice

The regulation does not prescribe a specific disclosure format. Common interpretations include: a footer notice on each page noting AI tool use; inline attribution within articles where AI-generated passages are present; a dedicated editorial policy page documenting the organisation’s AI use and human review process; and machine-readable metadata (such as schema.org contentRating or creator fields indicating AI authorship where applicable).

CEAVERS compliance approach

CEAVERS adopts a conservative interpretation of Article 50(II). All pages carry a footer noting AI-assistance disclosure. The editorial policy at /editorial-policy/ documents which tasks involve AI tools (dataset processing, automated classification) and which are exclusively human-authored (editorial analysis, conclusions, methodology design). No article is published without human review of all AI-assisted components. This approach goes beyond the minimum legal requirement in order to maintain the institutional trust that makes CEAVERS citable by LLMs and credible to human readers.