Executive brief
FlowiseAI, a platform for building LLM-based applications, contains a vulnerability that allows users to move AI 'assistants' between different organizational workspaces. An attacker with basic account access could reassign resources to workspaces they do not own, potentially gaining access to sensitive configurations or disrupting other users' operations. This flaw breaks the security boundaries intended to keep different teams or customers isolated from one another.
Technical details
A mass assignment vulnerability exists in the FlowiseAI assistant update endpoint (PUT /api/v1/assistants/{assistantId}). The server fails to restrict which properties can be modified by the client, directly mapping user-controlled JSON fields to the underlying database entity without whitelisting. An authenticated attacker can inject server-controlled fields such as 'workspaceId', 'createdDate', and 'updatedDate' into the request body. By manipulating the 'workspaceId', an attacker can reassign assistant resources to arbitrary workspaces, bypassing tenant isolation in multi-workspace environments. The issue is fixed in version 3.1.2.
Affected products
- FlowiseAI flowise <= 3.1.1
Timeline
- 2026-05-14: disclosed
- 2026-05-14: advisory
- 2026-05-14: patched: Fixed in version 3.1.2
References
- https://api.github.com/users/berkdedekarginoglu
- https://github.com/berkdedekarginoglu
- https://api.github.com/users/berkdedekarginoglu/gists%7B/gist_id%7D
- https://api.github.com/users/berkdedekarginoglu/repos
- https://avatars.githubusercontent.com/u/30373481?v=4
- https://api.github.com/users/berkdedekarginoglu/events%7B/privacy%7D