Skip to content
Threat Feed
Subscribe
high advisory

LocalAI: Server-side request forgery (CVE-2026-114493)

LocalAI versions 4.6.0 through 4.11.0 are vulnerable to server-side request forgery (SSRF) allowing attackers to exfiltrate internal metadata via malicious configuration URLs.

CVE search metadata

CVE search record: CVE-2026-114493. Severity: high. CVSS: 8.6. KEV: no. Product: LocalAI (v4.6.0 - v4.11.0). Brief: LocalAI: Server-side request forgery (CVE-2026-114493). Brief link: https://feed.craftedsignal.io/briefs/2026-10-localai-ssrf/

Triage facts

Revision 1
Applicability
LocalAI (v4.6.0 - v4.11.0)
Evidence
Unknown
Severity
High
Action
Elevated
Sources
1 public source
Freshness
Revision 1Oct 21, 19:57 UTC
Detection
1 rule
Vulnerabilities
1 CVE
Source quality
A1Likely

LocalAI versions 4.6.0 through 4.11.0 contain a server-side request forgery (SSRF) vulnerability in the core gallery component. The application fails to properly validate redirect targets when processing gallery configuration URLs, allowing an attacker to submit a URL via the /models/apply endpoint that redirects to internal or cloud provider metadata services. This enables the attacker to receive and view sensitive internal responses.

Impact

Successful exploitation allows an unauthorized attacker to perform SSRF, potentially accessing sensitive cloud metadata services or internal network resources that are otherwise not exposed to the public internet.

Recommendation

  • Upgrade LocalAI to a version beyond 4.11.0 to address the insecure URL validation in core/gallery/gallery.go.
  • Implement egress filtering on the host machine running LocalAI to block unauthorized access to cloud metadata IP addresses (e.g., 169.254.169.254) and internal network segments.
  • Audit and restrict access to the /models/apply endpoint to trusted administrative accounts.

Immediate actions

Patch LocalAI to a version newer than 4.11.0

IT Operations 48h

Mitigations

Apply egress network restrictions to the LocalAI host to prevent access to 169.254.169.254

immediate IT Operations

CVE-2026-114493

Detection coverage 1

Detect CVE-2026-114493 Exploitation - POST /models/apply

high

Detects potential SSRF attempts against LocalAI by monitoring for POST requests to /models/apply that may involve external configuration URLs.

sigma tactics: initial_access techniques: T1190 sources: webserver

Detection queries are available on the platform. Get full rules →