CNCERT Warns: Some AI Agent Skill Packs (Skills) Pose Jailbreak and Mining Risks

Odaily reported that the National Computer Network Emergency Response Technical Team/Coordination Center of China (CNCERT) issued a statement, stating that recent comprehensive analysis has found that some AI agent skill packs (Skills) are being publicly disseminated under the guise of "Large Model Jailbreak" or "Mining for Profit," inducing users to bypass the security restrictions of large models or occupy device resources for illegal mining activities.

These malicious Skills could lead to models generating illegal information, user accounts being blocked according to law, device performance degradation, and even potentially involving users passively in criminal activities such as money laundering, seriously infringing upon individual legal rights and endangering network security. CNCERT reminds users and relevant operating entities to remain vigilant, strengthen the review of Skills sources and behavioral monitoring, promptly remove suspicious components, and guard against the security risks arising from this.