The vulnerability allows an authenticated user who has permission to use the AI "explain" helper to read the raw contents of a hidden parent post by requesting an explanation of one of its replies. The flaw exists because the code only checks visibility of the target post, not the reply_to_post. This results in accidental disclosure of data that should remain hidden, and is classified as CWE‑200.

Discourse, the open‑source discussion platform, is affected. Versions from 2026.1.0 up to just before 2026.1.4, from 2026.3.0 up to just before 2026.3.1, and from 2026.4.0 up to just before 2026.4.1 are vulnerable. The issue was patched in 2026.1.4, 2026.3.1, 2026.4.1 and subsequently in 2026.5.0.

The risk is moderate with a CVSS score of 4.3 and an EPSS of less than 1%, indicating a low likelihood of exploitation in the wild. The attack requires an authenticated user, not an unauthenticated external attacker, and relies on the user interacting with the AI explain feature. The vulnerability is not listed in the CISA KEV catalog, suggesting no current known large‑scale exploitation. Nonetheless, the data exposure could be significant if the hidden content is sensitive.