// Security Advisory Triage

LangGraph SDK auth bypass in versions before 0.4.4

GitHub advisories — pip · published 2026-10-06 18:26 UTC
ACTION NEEDED PATCH_NOW SECURITYAPI
WHAT CHANGEDCVE-2026-104873 published 5 October 2026. LangGraph SDK versions before 0.4.4 silently ignore the actions= parameter on resource decorators when custom auth is enabled, allowing unauthorized operations. First patched version is 0.4.4.
WHY IT MATTERSAny custom auth implementation relying on actions= to restrict operations is not enforcing those restrictions. Users with read access can perform write, delete, or other restricted operations without detection or logging.
WHAT TO DOUpgrade langgraph-sdk to 0.4.4 or later. If you use custom auth with resource decorators, audit access logs for unauthorized operations that should have been blocked.
Model-reported confidence 95% (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-fvww-7h3r-vfhp
+summary: LangGraph SDK custom auth silently ignores actions= on resource decorators
+severity: high
+published_at: 2026-10-05T22:49:35Z
+cve_id: CVE-2026-104873
+vulnerabilities[].package.name: langgraph-sdk
+vulnerabilities[].first_patched_version: 0.4.4
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 reasoningLangGraph SDK auth bypass affects AI gateway and SDK security controls that platform leads must patch.
signalsAI SDK advisory · authentication/authorization bypass · custom auth silently ignores access controls · first patched version identified · high severity rating · known CVE assigned
suppression—
triage modelanthropic/claude-haiku-4.5 ($0.00000)
writer modelanthropic/claude-sonnet-4.5 ($0.00000)