A critical authorization flaw in OpenClaw allows sandboxed subagents to bypass security boundaries, potentially leading to unauthorized control over system processes and escalated execution policies.

The vulnerability is an incorrect authorization flaw (CWE-863) where the subagent control surface fails to validate request scopes correctly. A low-privilege authenticated user, acting as a sandboxed worker, can manipulate sibling processes and override tool policies.

The ability for a restricted process to escape its sandbox and gain control over sibling tasks poses a significant risk of lateral movement and privilege escalation within the application environment. Given the high CVSS score of 8.8, this vulnerability could allow an attacker to compromise the integrity and availability of the entire service, leading to potential data manipulation or unauthorized code execution with elevated permissions.

Immediate Action: Update the OpenClaw package to version 2026.3.11 or later to implement the necessary authorization checks.

Proactive Monitoring: Review system and application logs for unexpected subagent control requests or anomalous process termination events originating from low-privilege worker sessions.

Compensating Controls: Ensure that the host environment enforces strict Principle of Least Privilege for the service account running OpenClaw to limit the potential blast radius of a sandbox escape.

Public Exploit Available: No

Organizations utilizing OpenClaw must prioritize updating to version 2026.3.11 immediately to address this authorization bypass. Failure to patch leaves the system susceptible to privilege escalation attacks from within the sandbox, which is a critical security boundary for this product.