Kilde › Guides › AI Act Art 50 › AI-written articles: when the text dut…

AI-written articles: when the text duty bites, and what the editorial exemption really demands

Current to 26 August 2026 · updates land in the changelog.

Text is the quietly consequential corner of Article 50(4): AI-generated text published to inform the public on matters of public interest must be disclosed as artificially generated — a duty on the deployer, live since 2 August 2026, no transition.

The exemption, as the final guidelines define it

No disclosure is needed where the content has undergone human review and a natural or legal person holds editorial responsibility. The guidelines put teeth in both limbs: human review means “deliberate examination of the substance” by competent persons — spell-checks and cursory sign-offs are expressly insufficient — and editorial control means an entity with authority to approve, alter or reject the content and ultimate legal responsibility for it. Sharpest edge: any substantive AI edit after the review voids the exemption — the review must be the last substantive touch.

Who this catches

Newsrooms using drafting assistance are the obvious case and the best-equipped: a real editorial chain is exactly the exemption. The exposed cases are content operations without one — SEO farms, automated market updates, AI newsletters, corporate blogs on policy topics — where text goes out with no substantive human pass. “Matters of public interest” is the scope hook; treating news-shaped content as in scope is the defensible default.

Two boundaries that help

First, the assistive exclusions: AI translations, grammar work and short outputs sit outside the marking duty's reach by the guidelines' own examples — a human-written article run through AI translation is not an AI-generated article. Second, no retroactivity: pre-August-2026 archives do not need back-labeling. Note the machine-readable 50(2) marking of text is a separate, provider-side question — and text is the technically weakest marking modality, which is where the “technically feasible” qualifier earns its keep.

Related guides

Quick answers

Must AI-written articles be labeled in the EU?
AI-generated text published to inform the public on matters of public interest must be disclosed (Article 50(4)) — unless it passed human review of its substance and a person or entity holds editorial responsibility for it.
Does an editor's quick approval satisfy the exemption?
No. The final guidelines require deliberate examination of the substance by competent persons; cursory sign-off and spell-check-level review are expressly insufficient, and any substantive AI edit after review voids the exemption.
Ship the disclosures before enforcement finds the gap.

The AI Act Article 50 Kit: the disclosure copy library (English + 中文), the marking implementation guide with survival-test protocol, deepfake and text-labeling walkthroughs, obviousness and scope memos, the evidence log, the Code of Practice accession path — built from the regulation and the final guidelines, with pinpoint citations.

Get the kit — US$190 Free 4-page sample (PDF)

Instant download · 14-day unconditional refund · single-organisation licence · full product page

General information only — not legal advice, and no clearance opinions. Sources are Regulation (EU) 2024/1689 (Articles 50 and 99), Regulation (EU) 2026/1744, the Commission's final Article 50 guidelines of 20 July 2026 (non-binding) and our audited kit research. Marking-technology status changes fast: treat vendor announcements as live only once verified. © 2026 Kilde.

Built by Kilde's founder, a practising attorney admitted to a US state bar (not an EU or Hong Kong admission). About · Verification log · Refunds · Terms · Privacy · esau@trykilde.com