Cybersecurity researchers at Cyera disclose four vulnerabilities in OpenClaw that can be chained to enable data theft, privilege escalation, and persistence on a compromised system. The issues are collectively referred to as “Claw Chain.” According to the reporting, the flaws affect components of OpenClaw’s managed sandbox environment, including its OpenShell managed sandbox backend and its MCP loopback runtime. Researchers describe how an attacker could use the vulnerabilities together to first establish a foothold and then reach capabilities such as exposing sensitive information, escalating privileges, and planting backdoors that maintain continued control over the affected host. Both outlets characterize the risk as arising from chaining the four flaws rather than any single issue alone. The Next Web also notes that patches are available in OpenClaw, indicating that the vulnerabilities have been addressed. Overall, the disclosures focus on how weaknesses in the agent’s sandbox-related functionality could allow malicious activity that undermines the intended isolation boundaries, with the combined “Claw Chain” workflow supporting sustained, elevated access.