// Security Advisory Triage

LangGraph JavaScript custom auth permits cross-user Store access

GitHub advisories — pip · published 2026-10-31 05:45 UTC
ACTION NEEDED PATCH_NOW SECURITYAPI
WHAT CHANGEDGitHub published GHSA-4hm6-w6qq-w73v on 28 October 2026, a high-severity authentication bypass in langgraph-api. CVE-2026-118668. Custom authentication configurations may allow one user to access another user's Store data. Fixed in version 0.15.0.
WHY IT MATTERSAn authenticated user can read or modify data belonging to other users if your deployment uses custom authentication with LangGraph's Store feature. This is a cross-account data leak in a framework used to build stateful AI agents.
WHAT TO DOUpdate langgraph-api to 0.15.0. If you cannot update immediately, verify whether your deployment uses both custom authentication and the Store API — if it does not use Store, exposure is lower.
Model-reported confidence 85% (not measured accuracy) Primary source ↗ written by anthropic/claude-sonnet-4.5
Show the reasoning — raw diff, triage verdict, confidence
RAW DIFF
+ghsa_id: GHSA-4hm6-w6qq-w73v
+summary: LangGraph JavaScript custom authentication may permit cross-user Store access
+severity: high
+published_at: 2026-10-28T21:09:42Z
+cve_id: CVE-2026-118668
+vulnerabilities[].package.name: langgraph-api
+vulnerabilities[].first_patched_version: 0.15.0
HOW IT WAS JUDGED
Triage signals below are model-reported assertions, preserved verbatim; they are not independent arithmetic verification.
triage verdictMATERIAL at 0.92, needed 0.5
triage reasoningAuthentication bypass in AI SDK affecting data isolation between users requires immediate patching.
signalsAI SDK advisory · authentication bypass affecting cross-user data access · high severity · fixed version available (0.15.0) · LangGraph is widely used in AI agent frameworks
suppression—
triage modelanthropic/claude-haiku-4.5 ($0.00000)
writer modelanthropic/claude-sonnet-4.5 ($0.00000)