Discourse, an open‑source discussion platform, allowed attacker‑controlled filenames for uploaded files in chat messages to be rendered as raw, unescaped HTML within various excerpt components—such as chat channel lists, pinned message bars, email summaries, and preview buffers. This flaw permits trusted‑HTML injection: a malicious user could embed markup into a filename that would alter the visual presentation of excerpts. While the default Content Security Policy of Discourse blocks JavaScript execution and the advisory notes that no confidentiality, integrity, or availability impact has been demonstrated, the vulnerability still undermines the integrity of rendered content and could be leveraged by sophisticated attackers if CSP is weakened.

The flaw exists in all releases of Discourse preceding 2026.1.8, 2026.6.3, 2026.7.2, and 2026.8.0. Administrators running any of these problematic versions should consider them vulnerable. The affected product is the Discourse discussion platform as maintained by the Discourse community; no specific ecosystem is singled out beyond the platform itself.

The CVSS base score is 5.0, giving the feature a low‑to‑moderate severity rating. EPSS data is unavailable, and the vulnerability is not yet listed in CISA’s KEV catalog, suggesting it is not a high‑probability target for exploitation. The attack requires the attacker to be able to upload a file and send chat content, a condition that likely implies some level of account access; the primary attack vector is local or authenticated privilege. If a site disables the default CSP or otherwise relaxes HTML filtering, the risk of content alteration or script execution rises, potentially allowing phishing or defacement attacks. Current mitigations are therefore small‑scale structural changes rather than systemic compromise.