How the threat intel is made
AI-written summaries of public reporting, checked against their sources and refreshed together. Here is what that means for how far you should trust them.
What a briefing is
A briefing summarizes public reporting on one threat. It covers one of five kinds: actors and campaigns, exploited vulnerabilities, breaches and incidents, malware and ransomware, techniques and trends.
Every briefing answers the same six questions: what happened, who is behind it, what they want, what they did, who is at risk, what to do. Each answer comes twice. The plain version assumes no security background. The technical version is for whoever will hunt for it or patch it.
How it is written
An AI model writes each briefing from public reporting: vendor threat research, government advisories from CISA, the FBI and national cyber agencies, and established security journalism. A briefing needs at least two sources from different publishers before it can ship, and the build rejects one that has fewer.
Nothing here is original research, and nothing comes from private or client work. AI summaries can be wrong. A number gets copied from the wrong paragraph, a claim was later retracted, two groups get merged. The sources are there so you can check. Where something is disputed or claimed only by the attacker, the briefing says so under What is not settled.
How fresh it is
The section is refreshed as a whole, and every page shows the last refresh. Each briefing also has its own updated date. Pages warn you after 10 days without a refresh.
A briefing that first appeared today, by UTC date, is marked New until the next UTC midnight. A briefing whose activity stopped is marked historical or contained, not deleted. After 30 days with no reported activity it moves to the archive and keeps its page.
Following it
Every briefing ends in a revision history that says what changed and when, so you can tell whether anything you acted on has moved. The RSS feed lists briefings by their last update, so a changed one comes back to the top of your reader. The Patch now table on the index collects every CVE named in a live briefing, in the order to fix them.
What the labels mean
| Label | Meaning |
|---|---|
| Critical | Widespread or unauthenticated compromise happening now, or data exposure at very large scale. Act this week. |
| High | Active and damaging to the organizations it targets. Check whether you are one of them. |
| Medium and Low | Narrower targeting, lower impact, or mainly of interest for awareness. |
| High confidence | Several independent sources agree, including a primary one. |
| Moderate confidence | The core is reported by credible sources, but details are thin or rest on one vendor. |
| Low confidence | Early, contested, or claimed mainly by the attacker. |
| CISA KEV | Listed in CISA's Known Exploited Vulnerabilities catalog, which is evidence of exploitation. Not checked means nobody looked, which is different from not listed. |
Refreshed by: Claude routine intel-scout. The procedure and sourcing rules live in the repository's docs/INTEL_REFRESH.md.