OpenClaw versions before 2026.6.8 contain an authorization bypass vulnerability in the OpenAI‑compatible HTTP model override endpoint that allows callers with lower trust levels to perform actions normally protected by stronger authorization checks. The flaw, identified as CWE‑862 and CWE‑863, enables attackers to manipulate input so as to bypass administrative policies and execute restricted operations, potentially leading to unauthorized configuration changes or privileged function execution.

OpenClaw software, all releases prior to 2026.6.8 running on node.js environments. Systems that expose the HTTP model override capability to external callers without proper authorization controls are affected.

The CVSS score of 7.2 indicates a moderate‑to‑high severity. The EPSS score of less than 1% suggests the likelihood of exploitation is very low, and the vulnerability is not listed in CISA KEV, implying no known active attacks. The likely attack vector is a crafted HTTP request targeting the model override endpoint on a publicly reachable OpenClaw instance; a successful exploit would grant the attacker privileges normally reserved for administrators, enabling configuration tampering or other privileged actions. Although current exploitation risk is low, the weakness in the authorization model warrants prompt attention.