China's National Security Bureau warns of AI relay station risks
China's National Security Bureau issued a public warning about security and privacy risks from unregulated 'AI relay stations'—intermediary platforms that aggregate access to multiple AI models. The bureau identified four categories of risk: data interception and resale, model downgrading, malware implantation, and unauthorized overseas data transmission. The Central Cyberspace Administration launched a nationwide rectification campaign.
What happened
China's National Security Bureau issued a warning about unregulated 'AI relay stations'—third-party intermediary platforms that aggregate APIs from multiple AI model providers to give users access to various large language models through a single interface.
The bureau identified four major security risks: users' sensitive data may be stored on servers without proper encryption or access controls, and in some cases privately intercepted and resold; some relay stations substitute lower-capacity models for higher-end ones to cut costs; malicious actors may exploit hidden backdoors in relay stations to plant remote control malware on user devices; and some relay stations transmit user input data to foreign servers without required data-export compliance approval.
Reported impact
User privacy has been reportedly compromised through data interception and sale. Data transmitted abroad without proper authorization could result in exposure of personal information, trade secrets, and state secrets.
Index analysis
Relay stations have proliferated as demand for AI applications has grown, attracting users through convenience and cost savings. The market remains poorly regulated, with many operators lacking proper licenses and security safeguards.
Governance controls
The Central Cyberspace Administration of China launched a nationwide 'Clear Internet—AI Application Rectification' special campaign. The National Security Bureau advised users to select officially authorized, compliant platforms; apply anonymization to sensitive data before submission; rotate credentials regularly; and report suspected illegal activity through official channels.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.