CVE-2026-66005: janhq Jan CORS misconfiguration in local API server

CVE-2026-66005 · Severity: medium · CVSS 6.3 · Published 2026-07-24

Jan is an open-source AI desktop application that provides a local API server for running language models. A security flaw exists where exposing this server to a local network accidentally disables all access restrictions, allowing unauthorized users on the same network to use your AI models or read your data. This could lead to unauthorized use of computing resources or the exposure of sensitive information processed by the AI.

Technologies: Jan. Vendors: Jan.

Quick answers

Is CVE-2026-66005 exploited in the wild?
No. Junglewise Threat Intelligence has no report of CVE-2026-66005 being exploited in the wild.
How severe is CVE-2026-66005?
Medium severity, with a CVSS score of 6.3, published 24 July 2026.
Which products and versions are affected?
Jan through 0.8.4.
Is there a fix for CVE-2026-66005?
A patch was published on 2026-07-24: Fix merged in commit 3e1c1e7 Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.
How is CVE-2026-66005 attacked?
It is attacked over the network or from the adjacent network, in a network appliance.
Am I affected by CVE-2026-66005?
Check whether you run Jan at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

Technical details

A CORS misconfiguration and Host-header bypass exists in Jan's local API server (src-tauri/src/core/server/proxy.rs). When the server is bound to 0.0.0.0, the application logic replaces user-defined 'trusted_hosts' with a wildcard ('*'). This causes the 'is_valid_host' predicate to return true for any origin, leading the server to reflect arbitrary Origins with 'Access-Control-Allow-Credentials: true'. Attackers on the local network or those using DNS rebinding can bypass the unauthenticated OpenAI-compatible API to perform model inference, enumerate models, and read cross-origin responses. The vulnerability is fixed in commit 3e1c1e7 by enforcing the allowlist even when bound to all interfaces.

Affected products

  • janhq Jan through 0.8.4

Timeline

  • 2026-06-13: disclosed: Vulnerability reported to vendor
  • 2026-07-24: patched: Fix merged in commit 3e1c1e7
  • 2026-07-24: advisory

References

This page as Markdown. Free to reuse under CC BY 4.0 with a link back.