Executive brief
FastGPT, an AI application platform, contains a security vulnerability in its OpenAPI schema importer. An authorized team member could exploit this flaw to bypass internal security guards and access sensitive information from internal network services or cloud metadata. This could lead to the exposure of private infrastructure details or internal data.
Technical details
A Server-Side Request Forgery (SSRF) vulnerability exists in FastGPT's OpenAPI schema importer. While the application validates the top-level URL, it fails to apply the same internal-address guards to remote references ($ref) processed by the underlying SwaggerParser.bundle resolver. An authenticated attacker with team member privileges can provide a crafted OpenAPI schema that includes references to internal network addresses or cloud metadata services (e.g., 169.254.169.254). The parser fetches this content and returns it inline, allowing the attacker to read sensitive internal data. The issue is resolved in version 4.15.0-beta4 by implementing guarded fetching and rejecting non-local references during the bundling process.
Affected products
- labring FastGPT < 4.15.0-beta4
Timeline
- 2026-06-07: patched: Fix committed to repository.
- 2026-07-07: disclosed: CVE published.