Regarding the recent risks exposed by the open-source AI agent OpenClaw (commonly known as "Lobster") in terms of autonomous authority and data security, ByteDance's Volcano Engine has announced today that its cloud-based SaaS tool ArkClaw has been fully upgraded with an AI assistant security solution. This solution aims to build a multi-layered defense system from deployment environment to behavior permissions, transforming originally security-vulnerable open-source tools into compliant and controllable "digital employees."

image.png

The core evolution of ArkClaw lies in achieving deep isolation of the Agent runtime environment. Relying on cloud-native sandbox technology, all agent instances and third-party tools run within controlled containers, cutting off the path for single-point risks to spread to users' actual systems. At the same time, the system seamlessly integrates with Feishu bots, simplifying complex permission management into "least privilege" and "explicit authorization" mechanisms, ensuring that the AI assistant cannot access any protected sensitive resources without user's active approval.

In terms of dynamic control, Volcano Engine has introduced a full-process closed-loop protection mechanism. Before task execution, the system automatically intercepts high-risk commands such as file deletion through prompt intention recognition; during task execution, all suspicious network requests and system calls are monitored in real-time and stopped; after task completion, a complete operation trace is formed into an immutable audit log, allowing users to trace back at any time. This "high-risk operation review" mechanism effectively avoids the risks of data deletion or information leakage caused by the high autonomy of Agents.

Additionally, supply chain security has become a key focus of this upgrade. ArkClaw implements strict access scanning and regular inspections for all third-party Skills connected, ensuring the reliability of tool sources and preventing malicious plugins from being mixed in. By transforming AI from an "anonymous tool" into a digital assistant with a unique real-name identity and full behavioral traceability, Volcano Engine not only provides developers with a safer "breeding environment" but also establishes a solid security foundation for the large-scale deployment of generative AI in enterprise office scenarios.