Safeguard
Supply Chain•

Reviewed vs unreviewed advisories: why did my alert come late?

Some advisory databases only alert on reviewed records. Learn the difference, why a public CVE can stay silent, and how to cover the gap in your tooling.

Safeguard Security Team
3 min read

An alert can come late because many advisory databases separate reviewed records from unreviewed ones, and only reviewed ones trigger package alerts. A CVE that is public but still waiting for review can be invisible to the alert system even though the vulnerability is real.

Definitions

A reviewed advisory is one a database team has checked for validity and mapped to a package ecosystem, with a description and affected versions. An unreviewed advisory is a raw record imported from an upstream feed, with no such check.

GitHub's documentation states this plainly. Reviewed advisories cover supported ecosystems and are checked for a full description and ecosystem and package information. Unreviewed advisories are published automatically from the National Vulnerability Database feed and are not checked for validity or completion. Unreviewed advisories do not generate Dependabot alerts.

Why the split exists

Raw feeds are noisy. Some records have no package name, some name the wrong product, and some are later rejected. Alerting developers on every raw record would flood them with false matches. Review is the filter. The cost is delay, because review is a queue and a queue takes time.

GitHub also says new vulnerabilities may take time to appear in the advisory database and trigger alerts. The documentation gives no time window.

The gap in practice

Picture a CVE published on a Tuesday for a popular library. The record exists, but until it is reviewed and mapped, an alert tool that reads only reviewed data stays silent. A second tool that reads the raw feed might flag it at once, with some false matches mixed in. Neither is wrong. They made different trade offs.

How to cover the gap

  1. Know which sources each of your tools reads, and whether they alert on raw or reviewed data.
  2. Watch a small feed for high impact events, such as the KEV catalog, which lists vulnerabilities with confirmed exploitation.
  3. For critical dependencies, follow the maintainer's own security announcements.
  4. When a public CVE names a package you use and no alert has fired, check it by hand rather than waiting.
Source typeStrengthWeakness
Raw feedEarlyNoisy, poor package mapping
Reviewed databaseAccurate package dataReview delay
Maintainer announcementEarliest and most preciseScattered across projects

Where Safeguard fits

Safeguard's zero-day discovery work targets issues before a formal advisory exists, and its dependency scanning matches reviewed data against your inventory. A free scan shows what your project contains today.

For more on how records change once published, see advisories change after publication.

Quick check

Does an unreviewed advisory mean the vulnerability is not real?

No. It means it has not been checked and mapped yet. It may be real, wrong, or incomplete.

Should I turn on alerts for raw records?

Only if you can absorb the noise. Many teams prefer reviewed alerts plus a manual watch on the KEV catalog.

Sources

Last reviewed: 19 October 2026.

Never miss an update

Weekly insights on software supply chain security, delivered to your inbox.

Self-healing security runs on Safeguard.

Your first fix PR is minutes away.

No sales call required, even your agent can complete the purchase over MCP.