AI Science Tracker

Discourse post cooking can restore raw HTML inside a tag after sanitization

Discourse post-cooking and HTML sanitizer · AI involvement attributed to Anthropic OSS Scanner (item-specific model/version not identified)

Finding and attribution

While cooking a post, Discourse sanitized raw HTML while it was represented by a placeholder, then restored the raw HTML at the placeholder. A crafted placeholder inside a tag could therefore cause attributes to be inserted after the allow-lister had run. The project classified the fix as a security change; the reviewed primary source establishes HTML injection past the sanitizer, without stating a specific exploit outcome.

Product: Discourse post-cooking and HTML sanitizer. AI system: Anthropic OSS Scanner (item-specific model/version not identified). Team / organization: Anthropic and Discourse maintainers.

CVE identifiers: Not established in the cited evidence.

AI and human contribution (source account): The merged primary Discourse security pull request explicitly says “Discovered by Anthropic’s OSS Scanner” and identifies finding ANT-2026-ENYESJVJ. Anthropic’s OSS Scanner documentation states that findings are generated directly from its strongest models for this fast-track program; the item does not identify a specific model/version.

Human involvement is only described where the source account establishes it; unknown involvement does not mean none. Vendor confirmation of a flaw does not confirm every claim about AI contribution.

Discovery context: The project’s merged SECURITY pull request identifies a sanitizer-bypass HTML injection, directly credits Anthropic OSS Scanner, and records the fix. No CVE or GHSA appears in the reviewed primary sources, so the public PR #45728 is the case identifier. No JavaScript execution, exploitation, or zero-day claim is made.

Remediation status: fix documented. Discourse now drops any placeholder that occurs inside a tag, rechecking each substitution so hoisted raw HTML is restored only where text may appear. The change was merged in PR #45728 and backported to release 2026.10 (#45736), with related backports #45739 and #45740. Deployed patch status is not checked.

A source-documented fix does not establish installed patches or current safety. Source checking does not independently reproduce the finding or establish observed exploitation.

Disclosed 5 Nov 2026. AI attribution dated 5 Nov 2026. Sources checked 14 Nov 2026; these are separate dates, not a live installation check.

Recorded CVSS: Not available · no assessment established here. Documented technical effects: other.

Severity, consequences and exploitation evidence

Case grouping uses the highest recorded CVSS score, with its version-specific severity band. It is descriptive, not a combined or editorial risk index. All recorded assessments remain below; scores from different versions and sources may disagree. CVSS specifications define version-specific scoring and severity bands.

No source-backed CVSS assessment is established in this catalog.

Case-level evidence: other.
Prerequisites / affected scope: The Discourse patch report says raw HTML restoration could introduce attributes the allow-lister missed; no resulting browser action is stated..
Exploitation evidence: The cited source does not state whether this issue was exploited in the wild.. Impact source · checked 16 Nov 2026.

Documented fix: Discourse now drops any placeholder that occurs inside a tag, rechecking each substitution so hoisted raw HTML is restored only where text may appear. The change was merged in PR #45728 and backported to release 2026.10 (#45736), with related backports #45739 and #45740. Deployed patch status is not checked.

Original evidence

Download this source record (JSON)

Open this case in the interactive tracker · Browse all static security cases