Where the information comes from

We start with vulnerability records, government advisories, incident response teams, and established security reporting. We also use vendor advisories and project links when they appear in an official vulnerability record. Articles link to the sources behind their claims, including the original data for any scores we quote.

What goes into an article

Before drafting, the system collects what the sources say about who is affected, how the vulnerability works, what damage it can cause, and how to fix it. Each claim is saved with its source so it can be checked later. Advice on detecting attacks or checking a fix is included only when the sources support it.

How we check a draft

A separate AI model checks the draft against the collected evidence, looking for unsupported claims, contradictions, missing facts, and links that don’t support the text. Invented commands, tests, or explanations of how an attack works block publication. This is an automated check; it does not mean a person has reviewed every article or that we have reproduced the vulnerability ourselves.

When we publish

Drafts can publish automatically after passing both factual checks and an automated editorial review. Drafts that fall short get a revision. If the problems remain, they stay unpublished. If the checking service is unavailable, publication waits. A human reviewer cannot override a failed factual or safety check.

Who’s responsible, and what happens if we get it wrong

We publish under the name BlackShield Research and take responsibility for the articles. When we make a substantial correction, we update the article’s modification date. Changes to our writing process do not automatically rewrite older posts. If we no longer stand behind an article, we can remove it from the site and its feeds. We keep previous versions privately so we can review them before restoring a withdrawn article.