Technical impact
Zylon PrivateGPT’s URI Artifact Loader contains a flaw in the load_file_from_uri routine that allows an attacker to manipulate the Value argument and force the server to perform outbound HTTP requests. This vulnerability is classified as a server‑side request forgery (CWE‑918) and can result in exposure of internal networks, data exfiltration, or further lateral movement. The impact is confined to the server hosting PrivateGPT, yet it can affect any internal service reachable from that host.
Affected Systems
The issue affects Zylon PrivateGPT 1.0.0 and 1.0.1 deployments. The problem resides in the private_gpt/server/ingest/uri_loader.py component, specifically the load_file_from_uri function; no vendor‑issued patch or fix has been released as of the last update.
Risk and Exploitability
The vulnerability is exploitable remotely and does not require local access or privileged credentials. Although EPSS data is unavailable and the weak is not listed in the CISA KEV catalog, a publicly available exploit exists on GitHub, demonstrating that the attack can be automated against unpatched systems. The CVSS score of 6.9 reflects moderate severity, and the lack of mitigation from the vendor heightens the risk until a formal patch or administrative barrier is applied.