Chainlit versions through 2.12.0 allow authenticated users to delete or alter feedback entries belonging to other users by specifying arbitrary feedback identifiers in PUT and DELETE requests. This bypasses ownership checks and corrupts the human‑rated data that feeds model evaluation pipelines, leading to compromised data integrity and potentially misleading model performance assessments.

The vulnerability exists in Chainlit versions up to and including 2.12.0. All releases prior to the fix lack the ownership validation for feedback modification and deletion endpoints.

The CVSS score of 6 reflects a moderate threat level. EPSS data is not available, and the issue is not listed in the CISA KEV catalog, indicating no known active exploitation at present. The attack vector is authenticated use of the existing API; an attacker with valid credentials can manipulate or delete other users’ feedback records by supplying the target identifier.