The vulnerability is a SQL injection in the PM tag filtering function that enables bypassing tag filter conditions and accessing private message metadata that is not normally visible. Because of this flaw an attacker who can forge a request containing a crafted tag value can ask the application to return information about private conversations that it should not disclose. The weakness is an example of untrusted input reaching the SQL layer, leading to confidentiality loss as described by CWE-89.

Discourse, the open‑source discussion forum, is impacted in every release before 2025.12.2, 2026.1.1, and 2026.2.0. Versions 2025.12.2, 2026.1.1, and 2026.2.0 address the flaw and are recommended replacements.

The CVSS score of 4.9 reflects a moderate impact; the EPSS score of very low (<1%) and the absence from the CISA KEV catalog indicate that the vulnerability is not widely exploited. The likely attack path is via a web request that embeds a malicious tag value, enabling the execution of a crafted SELECT statement that returns private message metadata, leading to data disclosure. Because the exploit does not require local or privileged access, an attacker with internet connectivity to a Discourse instance can leverage the flaw.