FastGPT, an AI agent building platform, contains a Server‑Side Request Forgery vulnerability in the lafModule workflow node. The fetchData function uses axios to retrieve user‑supplied URLs without verifying if the target is an internal address, bypassing the isInternalAddress guard. Attackers who can access the workflow node, including unauthenticated users or any user with app editing privileges, can send arbitrary HTTP requests to internal or private network addresses, potentially exposing sensitive data or enabling further internal exploitation. This flaw corresponds to CWE‑918. The CVSS score is 2.3, indicating low overall severity but still a notable risk when combined with access privileges.

The affected product is labring:FastGPT. All installations running a version prior to v4.14.17 are vulnerable. Users running 4.14.17 or newer are not affected as the issue has been patched upstream.

The CVSS score of 2.3 and absence from the KEV catalog suggest that this vulnerability is not widely targeted or presently exploited. However, the likelihood of successful exploitation is non‑zero because the flaw allows unauthenticated or low‑privilege users to trigger outbound requests to internal destinations. An attacker can use this to discover internal hosts, read internal data, or pivot to other services. No public exploit is known yet, and the EPSS score is unavailable, so the exact risk can’t be quantified precisely. Nonetheless, because the flaw permits arbitrary network access from the server, it remains a potential attack vector worth addressing promptly.