CVE-2026-62186: Openclaw Auth Bypass Vulnerability

CVE-2026-62186 Overview

CVE-2026-62186 is an authorization bypass vulnerability in OpenClaw versions before 2026.6.8. The flaw resides in the OpenAI-compatible HTTP model override handling, where the application fails to consistently enforce authorization checks on incoming requests. Lower-trust callers can manipulate model override input paths to perform actions that require administrative privileges. The weakness is tracked as CWE-862: Missing Authorization.

The issue affects the openclaw:openclaw Node.js package and is exploitable over the network with low privileges and no user interaction.

Critical Impact
Authenticated attackers with low-trust credentials can bypass admin authorization policies and execute restricted operations against the OpenClaw HTTP interface.

Affected Products

OpenClaw (Node.js package) — all versions prior to 2026.6.8
Deployments exposing the OpenAI-compatible HTTP model override endpoint
Environments with multi-tenant or role-differentiated OpenClaw access

Discovery Timeline

2026-07-13 - CVE-2026-62186 published to NVD
2026-07-14 - Last updated in NVD database

Technical Details for CVE-2026-62186

Vulnerability Analysis

OpenClaw exposes an OpenAI-compatible HTTP interface that accepts model override parameters. The handler processes these parameters before evaluating whether the caller possesses the authorization required for the underlying operation. Because the authorization check does not cover every input path, a caller with limited privileges can drive the request into an execution branch that would normally require an administrative role.

The result is a broken access control condition. Operations gated behind admin policy — such as invoking restricted model routes or configuration-affecting actions — become reachable from lower-trust sessions. The vulnerability requires an authenticated context but does not require user interaction, making it usable in automated exploitation against exposed OpenClaw endpoints.

Root Cause

The root cause is missing authorization enforcement on a specific input path within the HTTP model override logic. The code trusts request-supplied override fields to select the downstream action without re-validating that the caller satisfies the policy for that action. This maps directly to CWE-862, where an actor can access a resource without being properly checked for permission.

Attack Vector

Exploitation is network-based against the OpenClaw HTTP service. An attacker holding valid low-privilege credentials crafts an HTTP request to the OpenAI-compatible endpoint and supplies model override fields that steer the handler into an admin-gated code path. Because the authorization decision is skipped for that path, the restricted operation executes under the attacker's session. Refer to the VulnCheck advisory and the OpenClaw GitHub Security Advisory GHSA-jhfx-v2j8-x3m6 for endpoint-level details.

No public proof-of-concept has been published, and the vulnerability is not listed in the CISA Known Exploited Vulnerabilities catalog.

Detection Methods for CVE-2026-62186

Indicators of Compromise

Requests to the OpenClaw OpenAI-compatible HTTP endpoint containing unexpected model override values from non-admin sessions
Successful invocations of admin-gated operations by accounts that lack administrative role assignments
Anomalous spikes in HTTP POST volume targeting model override parameters from a single low-privilege identity

Detection Strategies

Audit OpenClaw application logs for authorization decisions and compare the acting user role against the operation performed
Deploy Web Application Firewall (WAF) rules that flag or block requests where the model override field targets restricted routes from non-admin tokens
Correlate authentication events with subsequent privileged actions to identify role-to-action mismatches

Monitoring Recommendations

Enable verbose access logging on the OpenClaw HTTP layer, capturing caller identity, model override values, and resulting operation
Alert on any 2xx response to admin-scoped operations issued by accounts outside the administrator group
Monitor egress traffic from OpenClaw hosts for unexpected outbound calls that may indicate abuse of overridden model targets

How to Mitigate CVE-2026-62186

Immediate Actions Required

Upgrade OpenClaw to version 2026.6.8 or later, which contains the fix for the authorization bypass
Restrict network exposure of the OpenClaw HTTP interface to trusted networks and authenticated clients only
Rotate credentials and API tokens for any low-privilege accounts that could reach the affected endpoint

Patch Information

The vendor addressed CVE-2026-62186 in OpenClaw 2026.6.8. Details are published in the OpenClaw GitHub Security Advisory GHSA-jhfx-v2j8-x3m6. Node.js deployments should update the openclaw package dependency and redeploy affected services. Verify the installed version with npm ls openclaw after the upgrade.

Workarounds

Place OpenClaw behind an authenticating reverse proxy that enforces role-based access control on the model override endpoint before requests reach the application
Disable or firewall the OpenAI-compatible HTTP endpoint if it is not required for production workloads
Constrain accepted model override values to an explicit allowlist via proxy or gateway rules until patching is complete

bash

# Upgrade OpenClaw to the fixed release
npm install openclaw@2026.6.8
npm ls openclaw