Our policy on artificial intelligence
Our articles are produced with the assistance of language models. Here is exactly what that means, and what it does not mean.
What the models do
- they spot stories in the trade press, in publications and in public registers;
- they propose a selection, scored on genuine novelty, reach and verifiability;
- they write a first version following our template, with sources cited;
- they check facts, figures, similarity with sources, and the presence of required disclosures;
- they translate by adapting references, not word for word.
What the models do not do
- they do not publish on their own: a human validation precedes every publication;
- they do not invent a source, a quotation or an interviewee — an unverifiable quotation is removed;
- they do not sign under a fictitious journalist’s name: the byline is collective and owned;
- they do not decide what reaches Top stories.
Why we say so
Because it is verifiable, and because a newspaper hiding this would lose the only asset that counts: trust. If an error passes our checks, it is our responsibility, not a tool’s.