We use AI tools to research and write our coverage. This page explains how articles are produced and checked, what those checks cannot catch, how our bylines work and how we correct mistakes.
Who is responsible
Breach Signal is published by Breach Signal. Editorial responsibility: the editorial team of Breach Signal.
Articles are checked by automated tools and are usually published without a person reviewing each one first. We review published articles afterwards, act on reports from readers and correct errors openly.
How articles are made
- Software monitors a list of sources, such as official statements, trade publications and established news outlets, and skips stories the site has already covered.
- An AI research step collects information from several sources and records where each fact came from.
- An AI writing step produces an original article from those notes, without copying the sources’ wording.
- Automated checks look for copied text, compare names, numbers, dates and quotes with the sources, and score the article against an editorial checklist. Articles that fail are revised or dropped.
- A copy-editing step fixes grammar and style, and the article is published with a list of its sources.
The limits of these checks
The checks reduce mistakes but cannot catch all of them: AI tools can misread or misattribute information, leave out context or combine details incorrectly, and the sources themselves can be wrong, incomplete or later updated. Articles may therefore contain errors and may not be fully accurate, complete or current. Treat them as a summary and a starting point, and check the linked sources before relying on them.
Bylines
Our articles carry house bylines: pen names that Breach Signal uses for its beats. They do not refer to real people, have no personal credentials or experiences, and do not test products or give personal opinions as if they did. Each byline’s profile page describes the beat it covers and its writing style.
Sourcing
- Every article lists the sources it was based on, so readers can check them.
- News, explainers and analysis aim for at least three sources (two for short briefs), and key claims aim for two independent sources. Claims that cannot be matched to the sources are left out.
- Quotes are kept short, reproduced as written and attributed. Claims are attributed to whoever made them.
- Paywalled content is not used, and websites that ask automated tools not to visit them are respected.
What we do not publish
- Personal advice of any kind, including medical, legal, financial or security advice.
- Stories about private individuals, or allegations against identifiable people that are not attributed to a credible, named source.
- Personal data taken from leaks or breaches, or step-by-step instructions that would help someone cause harm.
- Product reviews or testimonials presented as personal experience, and fake comments or reactions.
Images
Images are licensed stock photos, illustrations made with AI and labelled “AI-generated illustration”, or simple title cards. AI-made images illustrate a topic and do not show real events or people.
Independence and advertising
Nobody pays for coverage, and advertisers have no say over what we publish. Advertising, if shown, is clearly labelled and kept separate from articles. Sponsored content or affiliate links, if we ever use them, will be labelled next to the content itself.
Corrections and feedback
When we get something wrong, we correct it openly with a dated note on the article, as described in our Corrections Policy. Questions, feedback and reports of errors are welcome through our contact page.