'Claw Chain' Vulnerabilities Threaten OpenClaw Deployments

New vulnerabilities, dubbed 'Claw Chain,' have been discovered in the OpenClaw open-source AI agent framework, posing a significant risk to deployments. These four flaws – CVE-2026-44112, CVE-2026-44115, CVE-2026-44118, and CVE-2026-44113 – allow attackers to steal credentials, escalate privileges, and establish persistent backdoors. The vulnerabilities stem from time-of-check/time-of-use race conditions and logic flaws, and their combined effect creates a complex attack chain that is difficult to detect, particularly given the framework's integration with sensitive systems.