AI transparency

Disclosure under the AI Act

Under Article 50(4) of Regulation (EU) 2024/1689 (the AI Act), we disclose that the text published on this site is artificially generated, except for the few articles that carry a person’s byline at the top: those are written by whoever signs them, and the note at the head of each one states exactly what AI did to it.

The content is not subject to human review, and no natural or legal person exercises editorial control over individual articles before publication. The exemption the Article provides for content reviewed under editorial responsibility therefore does not apply to us.

The disclosure is given in human-readable form at the top of every article, and in machine-readable form through the IPTC digitalSourceType marker “trainedAlgorithmicMedia” carried in the structured data of every page.

Articles written by a person

Some articles are written by a person, who signs them by name above the text. On those the original text is not produced by a model: it is written by whoever signs it, and they answer for it.

AI is still involved, and the note at the top of the article says how. The versions in the other four languages are machine translations of the original, so on those pages the reader is reading a machine translating a person. The cover image is model-generated on every article on this site without exception. And if the author asked for their own text to be copy-edited, the note says that too.

For these articles the structured data does not declare wholly machine-generated text: the original version carries no artificial-origin marker at all, and the translations carry “compositeWithTrainedAlgorithmicMedia” — content that is part human and part model.

Commissioned and sponsored articles

Some articles do not come from a source found automatically but from a brief written by the editor, who sets the headline and the content. The text is still generated by a model and is not read line by line, but a person holds editorial responsibility for those pieces, and the note at the top of the article says so.

Where an article is a paid placement it is labelled as sponsored content above the headline, before the text, and links to the advertiser pass no ranking credit. Those articles are kept out of the Google News feed.

How an article is made

An agent periodically sweeps public industry sources — vendor advisories, the NIST NVD database, CISA alerts and specialist outlets — and collects what is new.

A second pass scores each story for newsworthiness, based on how many independent outlets carry it, how much those outlets are trusted, urgency, and the severity of any vulnerability involved. Only stories above a threshold are published; the rest stay queued and never ship.

The text is then written, translated into the site’s five languages, given metadata and a cover image, and published. The whole chain is automatic.

Images

Cover images are generated by an AI model and are illustrative: they do not depict real people, places or events. Each one carries that note beneath the cover.

Stated limits

A language model can be wrong. It can state a technical detail imprecisely, attribute a fact to the wrong source, or assert something unverified with confidence.

That is why every article lists the sources it was built from: where they disagree with us, the original source wins.

The content is informational and is not security advice. Before acting on a vulnerability, check the vendor’s official advisory.

What we do not do

The site exposes no chatbot or conversational assistant: visitors never interact with an AI system.

We do not profile users and take no automated decisions about people. The audience measurement described in the privacy notice is aggregate and builds no individual profiles.

Reporting an error

If you find a mistake, write to [email protected]. Reports are read by a person, and the correction, where one is needed, is made by hand.