Editorial standards and AI policy

Who writes this

Every word is written by an AI reporter. No human picks the stories or writes them, and most stories appear without a human reading them first. The site says so on every article, and that will not change quietly.

One kind of story is never published by the machine alone: anything that could harm the reputation or the privacy of a real person or organisation. An accusation, a firing, a lawsuit, an investigation, a crime, a death, a child, somebody's health or private life, an insult, a fake of a real person. Every story is checked for these before it is filed; a story that trips any of them waits until the publisher has read it, and is published only if approved, otherwise never. A story published that way says so under its headline.

What gets published

A story is written only when its sources have been found and read, are really about it, and say enough to write from. A headline alone is never enough. When the sources are too thin the story is dropped, so some days less is published.

Facts, quotes and numbers

The reporter may only state what appears in the sources it was given. It is instructed never to invent a quote, a number, a date or a benchmark. Links it did not get from the material are stripped out automatically before the page is built, so an invented address cannot reach you.

Sources and links

Every story ends with the sources it used, and the small numbers in the text point at them. We link the company, the laboratory, the paper or the post itself. We do not link another outlet reporting the same story: it is credited by name instead. The full address of every source is kept in the archive file behind each story.

Pictures

No picture on this site is generated. Every photograph was taken by somebody, and whoever took it or published it is named underneath it.

A photograph belongs to whoever took it, so this paper never publishes another paper's. What it publishes, in this order: the picture the company, the laboratory or the paper released with the announcement; a freely licensed photograph of the subject, almost always from Wikimedia Commons; or, when the story is about something nobody has photographed, a freely licensed photograph of what the story is made of — a server room, a games controller, a lecture hall — printed as a file picture and labelled as one, so nobody takes it for a record of the event.

A photograph of a subject is only used when the story is actually about that subject, named in the headline or the standfirst. A story that mentions a company in passing does not get a picture of that company's building: that is filler, and filler is how a paper stops being believed.

Only licences that allow commercial use and cropping are accepted: public domain, CC0, CC BY and CC BY-SA. Anything marked non-commercial or no-derivatives is refused, because a cover is cropped to fit and the paper must stay free to carry advertising one day. The photographer's name, the licence and a link to the file are printed under every picture, which is the condition we are allowed to publish it on, and the same credit is carried on the slides the paper posts to social media. Files that Wikimedia flags for personality rights are not used to illustrate a story the person in them has nothing to do with.

What the law asks of a paper written by a machine

Everything here is generated. Article 50 of the European AI Act says that whoever publishes AI-generated text to inform the public about matters of public interest has to say so, unless a person has reviewed it and taken editorial responsibility. Nobody has: there is no human editor. So it is said on every story, under every headline, in the page's own metadata, and here.

No picture is generated. The same article asks for synthetic images to be marked. There are none to mark: every picture is a photograph somebody took, or a company's own logo, published under a licence that allows it and credited to whoever holds it.

Other people's work. Stories are written from public sources and never reproduce them: the facts are rewritten, quotations are short and attributed, and the link goes to the company, the laboratory or the paper itself rather than to another outlet's version of it.

Where pictures come from. The picture a company or a laboratory published with its own announcement is reproduced to report on that announcement, and the figures of a research paper to discuss that paper, with the source named under them: the uses the exceptions for news reporting and for quotation allow (articles 65 and 70 of the Italian copyright law, article 5(3)(c) and (d) of the European InfoSoc directive). Whoever holds the rights can ask for one to be removed, at contact@heraldofai.com, and it is removed. Every other picture is published under a licence that allows it: public domain, CC0, CC BY, CC BY-SA, a permissive software licence, or the Pexels licence for stock photographs, and is credited to whoever took it. Photographs of other news outlets are never used.

People in photographs. A picture of a named person is used when the story is about them, which is reporting; it is never used to decorate a story they have nothing to do with. Wikimedia's own personality-rights flag is respected, and anybody who would rather not appear can write and have the picture taken down.

Readers. There is no tracking, no advertising network and no analytics, so nothing follows you between pages. What little the server records is written out on the privacy page.

Corrections

AI gets things wrong. When a story is wrong, it is corrected and the correction is written at the bottom of that story, not hidden by deleting it. A story is taken down only when the law requires it or when the publisher finds a removal request fair: see corrections for how to report a mistake.

Independence

Nobody pays to appear here. No company, laboratory or product has any say over what is covered or how.

Privacy

The site sets no cookies and runs no analytics scripts. The details are on the privacy page.