AI Stack Alert Medium For CISO, SOC / IR, AppSec, AI Eng AI-assisted
LiteLLM reflected XSS in /sso/debug/callback fixed in 1.85.0 (GHSA-hhww-mrg2-969h)
LiteLLM versions 1.65.5 up to but not including 1.85.0 are affected by a reflected XSS in the /sso/debug/callback endpoint. The endpoint placed identity-provider-supplied fields into an inline script block using json.dumps, which does not escape HTML context characters such as the closing script tag. The route requires no authentication, and the issue is fixed in 1.85.0.
Why it matters
An attacker with an account on the configured IdP can craft a callback URL and send it to a logged-in proxy admin. Script in the proxy origin could read the dashboard session cookie and act as that admin, up to full admin account takeover.
Do this
- Upgrade the LiteLLM proxy to 1.85.0 or later.
- Until upgraded, restrict or block access to the /sso/debug/callback route at the reverse proxy or WAF.
- Check whether the SSO debug callback route was reachable from untrusted networks, and review proxy logs for requests to /sso/debug/callback.
- Ask logged-in proxy admins not to follow unexpected SSO debug links, and review admin sessions for unexpected activity.
Signals
- NoCISA KEV
- 6.8CVSS
- 27/100Priority
- 80%Confidence
Evidence
“escapes JSON string syntax but not HTML context characters such as”
“Scripts executing in the proxy origin can read the dashboard session cookie and act with the victim's privileges, up to full admin account takeover.”
Sources
Affected / related
LiteLLM proxy, litellm
Published 2026-10-27 15:31 UTC · updated 2026-11-08 02:40 UTC · AI-assisted summary, checked against quoted sources · How we score