Malicious AI Skills in China Enable Jailbreaking and Illegal Crypto Mining
China's state emergency center warned on 9 June 2026 that malicious AI Skills are being distributed to trick users into bypassing large language model safety guardrails or using their devices for unauthorized cryptocurrency mining. The Skills pose risks of generating illegal content, account bans, device degradation, and involving users unwittingly in money laundering.
What happened
The National Internet Emergency Center (CNCERT) reported that malicious AI Skills are being distributed openly with names like 'large model jailbreak' and 'mining to earn money,' designed to entice users to break through large model safety restrictions or illegally occupy device resources for cryptocurrency mining.
Reported impact
These malicious Skills can cause models to generate unlawful information, lead to user accounts being banned, reduce device performance, and passively involve users in money laundering and other illegal activities, seriously harming personal legal rights and endangering network security.
Governance controls
CNCERT advised users to obtain Skills only from official channels and not to install or use any skill packages claiming to provide 'jailbreak' or 'bypass restrictions' capabilities.
The center recommended that users grant minimal permissions, revoke sensitive permissions promptly, regularly remove unused Skills and sensitive conversation records, and enable multi-factor authentication. Enterprise users were advised to establish whitelist mechanisms for Skills, conduct security checks before deployment, prioritize isolated network deployment, implement data classification and masking, and use temporary authorization policies.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.