Flowise CVE-2026-56270

The vulnerable Flowise instance (<=3.0.13) must be network-reachable to the attacker, the /api/v1/loginmethod endpoint must be exposed (true in default configurations including FlowiseAI Cloud and standard self-hosted deployments), and the attacker must supply a valid organizationId UUID for a tenant that has at least one SSO provider configured with non-empty credentials. No authentication, no user interaction, no session, and no special platform are required. The only meaningful limiting factor is knowledge of an organizationId - guessing a random UUIDv4 is infeasible, but IDs frequently leak via invite links, public dashboards, browser history, support tickets, or enumeration of related endpoints; tenants that have never configured SSO secrets are not exploitable even though the endpoint still responds.

Real-world risk is high for any internet-exposed Flowise tenant that has configured SSO. The CVSS 4.0 vector AV:N/AC:L/AT:N/PR:N/UI:N is consistent with the description - a single unauthenticated HTTP GET - and the PoC in the GHSA confirms trivial reproducibility. Confidentiality impact is rated H (VC:H) while integrity and availability are N, which accurately captures the bug as a pure secret-disclosure flaw; however, the downstream blast radius is larger than VC:H alone suggests because leaked Google/Microsoft/GitHub/Auth0 client secrets can be combined with the corresponding client IDs to impersonate the tenant's OAuth application, mint tokens, or pivot into the identity provider. There is no CISA KEV listing and no EPSS score in the input, so widespread mass-exploitation is not confirmed, but publicly available exploit code exists and the only attacker prerequisite is a valid organizationId (a UUID that may be obtainable via enumeration, error messages, public invite links, or leaked URLs). One ambiguity worth noting: the CVSS vector indicates AV:N/PR:N, but exploitation in practice still requires the attacker to know or guess an organizationId - guessable but not zero-friction.

An attacker enumerates or obtains a tenant's organizationId - for example from a leaked invite URL, an error response, or by scraping it from a public Flowise instance - then issues a single unauthenticated GET request to https://target/api/v1/loginmethod?organizationId=<uuid>. The 200 OK response contains the tenant's SSO provider records with cleartext clientSecret values, which the attacker then uses with the matching clientID to mint OAuth tokens against Google/Microsoft/GitHub/Auth0 and pivot into the victim organization's identity provider. A working PoC is published in the vendor's GHSA advisory.