BEIJING - A Beijing-based research team has introduced a safety detection tool called Foresight ClawAudit on Wednesday and made it freely available to help users ensure safety while using the open-source AI agent OpenClaw, Beijing Evening News reported on Thursday.

Jointly released by the Beijing Institute of AI Safety and Governance and the Beijing Key Laboratory of Safe AI and Superalignment, the Foresight ClawAudit can scan over 30 checkpoints across eight security types. It includes built-in detection for more than 270 security vulnerabilities related to OpenClaw and is continuously updated to enhance its effectiveness, according to the report.

"At present, this tool primarily provides users with risk detection and notification. And users need to decide whether to address the identified risks based on their specific application requirements," Zeng Yi, director of Beijing Institute of AI Safety and Governance, was quoted as saying in the newspaper, adding that necessary automated risk remediation features will soon be launched to offer users more options.

READ MORE: Security calls for safe-clipping 'lobster' claws

The Foresight ClawAudit will continue to introduce and improve the types and coverage of security detection for agent frameworks, said Zeng.

Meanwhile, he cautioned that no single security detection tool can address every potential security risk. Their goal is to mitigate these risks to the greatest extent possible.

READ MORE: Platforms tighten rules on AI-generated content

China's Ministry of Industry and Information Technology has recently issued a warning that some OpenClaw-powered deployments carry high security risks when in default or improper configuration, making them highly susceptible to cyberattacks, information leakage, and other security issues.