How the Vexday Newsroom works

Vexday news is written automatically from public sources and checked by rules before it goes live. This page explains how each story is made, what the artificial intelligence does and never does, and how to request a correction.

Who writes

"Vexday Newsroom" is the name of the automated system that writes the news. No journalist signs the stories and no human reviews each text, and the page never pretends otherwise: every story ends with a notice that it was written automatically.

The focus is cybersecurity, with attention to Brazil and Latin America. For now, stories are published in Portuguese.

Where the news comes from

  • Dozens of public sources: specialised outlets, government agencies and CERTs, vendors and researchers.
  • Articles from different outlets about the same event, even in another language, become a single story.
  • The artificial intelligence reads the sources and separates the facts: what happened, who says each thing and with which excerpt of the source. A fact that is not in the source is discarded.

What goes live on its own and what does not

  • Level A: the official source itself confirms it (the vendor, the public agency, the affected organisation). The story states what was confirmed.
  • Level B: an outlet's report. Everything is attributed ("according to outlet X"), and the story says the information is preliminary.
  • Level C: a claim made only by criminals (such as a ransomware group), an unconfirmed leak, an accusation, personal data, a physical threat, unconfirmed attribution of an attack and conflicting sources. None of these is published on its own: they go to human review or to Vexday's investigation.

The check before publication

Every number, CVE, company, product or group name and every link in the text must be in the sources' facts. A text that fails this check does not go live. The artificial intelligence does not search the internet to write: it only uses facts that were already checked.

The feature (Intelligence News)

Some stories get a longer feature, on the same page. That only happens when Vexday has its own data to add: the CVE in CISA's exploitation catalogue or with a known exploit, a group's history in the Vexday database, victims in Brazil. Without that, the feature would just rewrite someone else's report, and the short story stays. A criminal claim never becomes a feature.

Sources

The outlet that broke the news is named, with a link to the original report. Aggregators and specialised databases appear by type ("a leak-site aggregator").

Updates, corrections and withdrawal

  • A story changes on the same page when a fact really changes, and the timeline shows each version.
  • A correction appears on the story itself, with its date.
  • When new information contradicts the text, the page shows a review notice until the decision: keep, correct or withdraw.
  • A withdrawn story stays at its address, with a notice that it was withdrawn.
  • When two published stories are the same event, one points to the other.

Images

Some features carry an AI-generated illustration, always with a notice: it has no evidentiary value and does not show real people, companies or systems.

How to request a correction or withdrawal

Write to contato@vexday.io with the story's address, what is wrong and, if available, the public source that shows the correct information. Organisations named in a story can ask for the text to be reviewed at the same address.

Every request is reviewed. When it holds, the correction appears on the story itself, with its date; when the story does not hold up, it is withdrawn.

See the news