Openclaw cli: OpenClaw Canvas Authentication Bypass Vulnerability (CVE-2026-3690)

The vulnerability in OpenClaw CLI (CVE-2026-3690) is due to the `authorizeCanvasRequest()` function implementing an IP-based fallback authentication for canvas endpoints. When a WebSocket client authenticates from a private IP, all subsequent HTTP requests from that IP are granted access without verifying individual authentication tokens. This is problematic in environments where multiple clients share an IP, such as corporate NATs, VPNs, Kubernetes clusters, or Docker host-mode networking. The root cause is that the IP-matching function only checks if any authenticated WebSocket client exists for the IP, without verifying user/session identity. Canvas endpoints use this weaker authentication path instead of the standard token-based method. An attacker on the same private network can exploit this to gain unauthorized canvas access. The vulnerability affects versions before 2026.7.1. A fix is available.

An attacker sharing the same private IP address as an authenticated WebSocket client can bypass authentication and gain full access to OpenClaw canvas endpoints, including the A2UI interface, canvas content, and WebSocket connections. This compromises confidentiality and integrity of canvas data. The vulnerability does not affect public IPs and requires network proximity or shared IP environments such as corporate NATs or container networks.

A patch is available for OpenClaw CLI to fix this authentication bypass vulnerability. Users should upgrade to version 2026.7.1 or later. Until patched, restricting network access to trusted clients and avoiding shared IP environments for OpenClaw canvas endpoints may reduce risk, but the definitive solution is to apply the official fix.