Impact
LocalAI versions 4.6.0 through 4.11.0 contain a server‑side request forgery vulnerability where the gallery configuration endpoint accepts a POST with a configuration URL and validates only the first location. When the URL redirects, the redirect target is fetched and its content is returned to the requester, bypassing the SSRF guard. The consequence is that an attacker can cause the application to send requests to internal or cloud metadata services and read their responses, exposing sensitive system data and potentially credentials.
Affected Systems
The vulnerability is present in the LocalAI:LocalAI application. All releases from 4.6.0 up to and including 4.11.0 are affected.
Risk and Exploitability
The CVSS score of 9.2 reflects a high‑risk. EPSS is not available and the vulnerability is not currently listed in the CISA KEV catalog, but the lack of monitoring does not diminish exploitation potential. Attackers can exploit the flaw remotely by sending a crafted POST request to the /models/apply endpoint. The exploitable target must be reachable by the LocalAI host, meaning that systems exposed to untrusted input through that endpoint are at risk of internal data disclosure.
OpenCVE Enrichment