Executive brief
Flowise, an open-source tool for building LLM-based applications, contains a security flaw where it inadvertently includes sensitive user data in its API responses. When a user logs in or is invited to the platform, the system returns the encrypted version (hash) of their password to the requester. An attacker could collect these hashes and use specialized hardware to attempt to crack them, potentially gaining full access to user accounts and the sensitive AI workflows they manage.
Technical details
A sensitive information disclosure vulnerability exists in FlowiseAI Flowise (up to 3.0.12) within the API Response Handler. Specifically, the `login()` and `saveInviteAccount()` functions in `packages/server/src/enterprise/services/account.service.ts` fail to sanitize user objects before returning them to the client. This results in the `credential` field (containing the bcrypt password hash), `tempToken`, and `tokenExpiry` being included in the JSON response body of the `/api/v1/account/login` and `/api/v1/account/invite` endpoints. While the attack complexity is rated as high due to the requirement of specific API interactions, a remote attacker can harvest these hashes for offline brute-force or dictionary attacks. This issue is a regression or incomplete fix of a previous security patch (PR #5167) that introduced a `sanitizeUser()` function but failed to apply it to all relevant authentication pathways. Users are advised to upgrade to a version beyond 3.0.12.
Affected products
- FlowiseAI Flowise <= 3.0.12
Timeline
- 2026-03-05: other: Vulnerability discovered and PoC created by researcher
- 2026-05-06: disclosed: NVD and VulDB disclosure
- 2026-05-06: advisory: GitHub Advisory published