GoPlus Security Introduces AgentGuard for AI Agent Protection

GoPlus Security has launched a security tool named AgentGuard, designed specifically for AI agents with real execution capabilities. According to Foresight News, this tool aims to provide execution-level security protection for OpenClaw. As AI agents begin to perform real execution tasks such as command execution, file reading and writing, service calls, and on-chain transactions and signatures, the risks have escalated from incorrect responses to potential system damage, data leaks, or financial losses. Leveraging its extensive experience in Web3 on-chain security and transaction risk control, GoPlus has developed a proactive security measure to address these risks before AI execution occurs. AgentGuard employs a mechanism of 'runtime interception and on-demand deep scanning' to intercept, confirm, or allow high-risk actions, assessing the risk of actions before AI agents execute them. This execution-level security mechanism is specifically designed to prevent AI agents from executing unauthorized, dangerous, or inappropriate actions, ensuring safer use of OpenClaw. AgentGuard is now open source and available for AI builders and developers.