Clickout Media sacked its reporters but kept their bylines on AI articles

When a byline becomes a prop

Clickout Media, a UK publisher running local news sites, made around 20 journalists redundant in November 2024 ^[1]. What happened next is the strange part. Articles began appearing on its sites under the bylines of people who no longer worked there, written not by them but by AI ^[1].

One former reporter, Scott Latham, discovered his name above an article he had never written ^[1]. The piece was about a court case. It read like something a machine might produce if asked to summarise a handful of facts without much care. Latham hadn't seen it, approved it or been asked.

Think of it like finding your name on a takeaway menu after you've left the restaurant. The food is real, customers order it, but you had nothing to do with the kitchen. Except here, what was being served was information, and people reading it thought they knew whose reporting they were trusting.

What the AI Act says about disclosure

Article 50 of the AI Act sets out what has to be disclosed when AI-generated content reaches the public ^[2]. Since 2 August 2026, AI-generated text published to inform the public on matters of public interest must be disclosed as such, unless a human has reviewed it and someone holds editorial responsibility ^[2].

The rule doesn't ban AI-written journalism. It asks for honesty: if a piece is written by AI and no human has properly reviewed it and stood behind it, say so. The moment someone does review it and take responsibility, the disclosure duty falls away.

Clickout Media told Press Gazette the articles were edited by humans before publication and labelled as AI-assisted ^[1]. Former staff said they saw no such labels, and the company didn't explain who those editors were or whether they had taken editorial responsibility in the sense the regulation describes ^[1].

Why bylines matter

A byline is more than decoration. It tells the reader who did the work and who to hold to account if the facts are wrong. Readers don't fact-check every sentence. They judge whether to trust a piece partly by the name above it.

When that name belongs to someone who left months ago and had no part in writing, checking or approving the article, the byline becomes misleading. It suggests a human reporter gathered the facts, made editorial judgements and stands behind the result. If none of that happened, attaching the name is a problem.

The AI Act's transparency rule is built on a similar idea: people have a right to know what they're dealing with. If the content is AI-generated and nobody with a name and a desk has vouched for it, the public should be told ^[2].

The evidence question nobody asked

When former staff challenged Clickout Media, the question became: what did the company actually publish, when, and who approved it? That's an evidence question.

The company said the articles were labelled and edited. The journalists said they saw no labels and had no involvement. In a dispute like this, it matters who can show what was published, in what form, at what time.

Screenshots can be dismissed as selective or edited. Cached pages disappear. If the publisher had sealed each version of each article as it went live, together with any disclosure text and the name of the person who approved it, the dispute would be shorter. Anyone could check that the sealed record matched what readers saw, unchanged.

How a witness would work here

The thing to seal is each article as published: the text, the byline, any AI disclosure and the approver's name. Sealing happens at publication.

The article itself stays on the publisher's server. What gets sealed is a fingerprint. Anyone the publisher shares the sealed record with, a regulator or a former employee, can recalculate the fingerprint from the published version and check it against the sealed one. If they match, nothing has changed since the article went live. Change a single word and the fingerprints won't match.

A sealed record shows what was published and when. It doesn't show that the article was accurate, that the byline was used fairly or that the disclosure was adequate. Those are separate questions. But it does give both sides a fixed point to argue from, rather than a swamp of screenshots and denials.

The first step is to decide who in the organisation approves publication and to seal a record of that approval at the same time as the article. The content never leaves your systems. You send the fingerprint to WitniumChain through the REST API, the SDK or the MCP server, and get a witness ID back. The first 2,000 witnesses a month are free. Read the full documentation to see how it fits your workflow.

How it would work

1. At publication, seal a fingerprint of the article (text, byline, disclosure text) and a record of who approved it.
2. Share the sealed record with anyone who questions what was published.
3. They recalculate the fingerprint from the live or archived version and compare it with the sealed one.

What this means for you

If you publish content, AI-generated or otherwise, ask:

• Who in your organisation approves each piece before it goes live, and is that person's role clear?
• If someone questions what you published six months ago, could you show exactly what appeared, with whose approval?
• If you use AI to draft, summarise or generate articles, have you decided when and how to disclose that, in line with Article 50?
• If you use bylines, do the people named know their names are being used, and do they stand behind the content?

Transparency rules are easiest to follow when you have a clear record of what you did. Sealing that record as it happens means nobody has to trust your memory or your filing system.

What to watch

Article 50 of the AI Act has applied since 2 August 2026 ^[2]. Penalties for breaching transparency duties can reach €15 million or 3 per cent of global annual turnover, whichever is higher, though for small and medium enterprises the lower figure applies ^[3].

The UK is developing its own AI rules, but transparency about AI-generated content is likely to remain a theme. Publishers elsewhere in Europe face the same question Clickout Media now faces: when AI writes and a human name goes above it, what do readers deserve to know?