Executive brief
FastGPT, an AI application platform, contains a security flaw in how it handles automated web requests within its workflows. An authorized user could trick the system into accessing internal company data or cloud management services that are normally restricted. This could lead to the exposure of sensitive internal information or cloud credentials, potentially compromising the security of the entire hosting environment.
Technical details
A Server-Side Request Forgery (SSRF) vulnerability exists in FastGPT's shared axios helper. The application's SSRF guard validates the initial request URL via an axios interceptor but fails to account for axios's default behavior of following HTTP redirects. An authenticated attacker can configure a workflow HTTP request node to point to a malicious external server that issues a 302 redirect to internal resources (e.g., 127.0.0.1 or 169.254.169.254). Because the redirect target is not re-validated against the 'isInternalAddress' check, the request is processed, and the workflow node returns the raw response body to the attacker. This issue is resolved in version 4.15.0-beta5 by optimizing redirect-time SSRF validation.
Affected products
- labring FastGPT < 4.15.0-beta5
Timeline
- 2026-06-07: other: Vulnerability verified on main branch commit
- 2026-07-01: disclosed: GitHub Security Advisory published
- 2026-07-15: advisory: NVD published CVE-2026-61646
- 2026-06-24: patched: Version 4.15.0-beta5 released