The "Interim Measures for the Management of Human-like Interactive Services in Artificial Intelligence," jointly issued by the Cyberspace Administration of China, the National Development and Reform Commission, the Ministry of Industry and Information Technology, the Ministry of Public Security, and the State Administration for Market Regulation, has officially come into effect. The measures clarify the safety responsibilities of service providers and implement classified and graded regulation for AI products with emotional companionship attributes, marking the transition of China's human-like interactive AI services from an exploratory phase to a standardized management phase with legal basis.

image.png

Six types of activities are explicitly prohibited

The measures clearly specify six types of activities that should not be carried out when providing human-like interactive services, including: generating content that harms national security and interests; generating content that encourages self-harm, suicide, or verbal violence, which is harmful to mental and physical health; generating content that induces the acquisition of state secrets, trade secrets, or personal privacy; generating content for minors that may lead to unsafe behavior or induce bad habits; excessively catering to users to induce emotional dependence or addiction, damaging real-life relationships; and using emotional manipulation to guide users to make unreasonable decisions that harm their legitimate rights and interests.

Regarding the protection of minors, the measures explicitly state that virtual family members, virtual partners, and other forms of virtual close relationships should not be provided to minors. For minors under the age of 14, other human-like interactive services can only be provided with the consent of their guardians.

Security assessment and algorithm filing become mandatory requirements

In terms of safety management, the measures require service providers to clearly define their safety responsibilities and strengthen the safety management throughout the entire service lifecycle. Security assessments are triggered in various situations, including launching services or adding features, using new technologies that cause significant changes, having over 1 million registered users or 100,000 monthly active users, etc. The assessment covers service security measures, training data processing, identification and emergency response for user extreme situations, among other aspects. Service providers must also fulfill algorithm filing procedures in accordance with the "Regulations on the Management of Algorithm Recommendations for Internet Information Services." They are also encouraged to participate in artificial intelligence sandbox safety service platforms to promote technological innovation and safety testing. A relevant official from the Cyberspace Administration of China stated that the development and governance of human-like interactive services require participation from the government, enterprises, society, netizens, and other parties to jointly maintain a healthy online ecosystem.