Executive brief
Chainlit is a platform for building conversational AI applications. The platform fails to properly verify that users own feedback records when modifying or deleting them. Authenticated attackers can corrupt or delete other users' feedback data, which undermines the integrity of human-generated ratings used to evaluate and improve AI models.
Technical details
Chainlit versions through 2.12.0 lack proper ownership validation in the PUT and DELETE endpoints that handle feedback records. The vulnerability is an authorization bypass: while authentication is required to access these endpoints, the application does not verify that the authenticated user is the owner of the feedback record they are modifying or deleting. An attacker with valid credentials can supply arbitrary feedback identifiers to alter or remove feedback belonging to other users, corrupting evaluation datasets. The fix requires adding ownership checks before allowing modifications to feedback records.
Affected products
- Chainlit Chainlit through 2.12.0
Timeline
- 2026-08-28: disclosed