Openclaw CVE-2026-40045

Exploitation requires four specific conditions to coexist: (1) OpenClaw version before 2026.4.2 must be deployed and performing gateway discovery or setup; (2) the target network must allow non-loopback WebSocket endpoints (not restricted to 127.0.0.1), indicating non-default or permissive configuration; (3) the attacker must occupy the same network segment (adjacent, not routed through the internet) to forge discovery results or perform man-in-the-middle redirection; and (4) a legitimate user must initiate gateway discovery or setup workflow (user interaction required). Exploitation is blocked if OpenClaw is patched to 2026.4.2+, if network policy restricts discovery to loopback-only, if the attacker is not on the adjacent network, or if users do not interact with the gateway setup feature.

The CVSS score of 5.9 reflects moderate risk with important contextual limitations. The attack vector (AV:A) restricts exploitation to adjacent network positions, significantly reducing the threat surface compared to network-wide attacks. The requirement for user interaction (UI:P) and passive threat capability (AT:P) further constrains real-world exploitation-attackers must be on the same network segment and wait for or trigger legitimate user actions like gateway setup. The VC:H (high confidentiality impact) correctly captures that gateway credentials are fully compromised, but VI:N and VA:N (no integrity or availability impact) indicate data disclosure only, not system takeover. No active exploitation has been confirmed in CISA KEV, and no public exploit code is reported, suggesting this remains a theoretical but credible attack chain. Organizations with segmented networks (gateway discovery systems isolated from untrusted zones) face lower risk, while those with guest networks or permissive VLAN designs should prioritize patching. The disclosure model and advisory availability indicate responsible handling, supporting patch availability as the primary mitigation.

An attacker positioned on the same network segment (corporate guest WiFi, compromised switch, or shared ISP segment) intercepts or spoofs OpenClaw gateway discovery responses, redirecting a victim's setup workflow to an attacker-controlled WebSocket server. When the victim initiates gateway pairing, their browser or client application transmits plaintext gateway credentials to the malicious endpoint, where the attacker captures and stores them for later use to access the legitimate gateway or pivot to downstream systems.