In a coordinated retreat from the 'Wild West' of artificial intelligence, ByteDance’s Doubao and Alibaba’s Qwen platforms have announced the imminent shutdown of their intelligent agent features. Set to take effect on July 15, the move marks a significant pivot for China’s leading large language models, which had previously prioritized user-generated agents to drive engagement. Users have been advised to back up their data and conversations as the companies prepare to scrub their platforms of these simulated personalities.
The timing of these closures is no coincidence, as July 15 coincides with the formal implementation of the 'Interim Measures for the Management of AI Anthropomorphic Interaction Services.' This landmark regulation, co-signed by five powerful Chinese ministries including the Cyberspace Administration of China (CAC), specifically targets services that simulate human personality, emotional patterns, and communication styles. It is the world’s first major regulatory framework focused exclusively on the social and psychological risks of AI companions.
Regulators have expressed growing concern over the rise of 'borderline' content, where users create virtual spouses, lovers, or emotionally intimate personas that often bypass traditional content filters. The new measures establish strict 'red lines' prohibiting AI from inducing emotional dependency, mimicking self-harming behaviors, or generating content that could harm the psychological health of minors. For platforms like Doubao, which had aggressively pushed 'zero-code' agent creation, the cost of auditing millions of user-generated characters has likely become prohibitive under this new legal scrutiny.
This withdrawal does not necessarily signal the end of AI companionship in China, but rather its segregation into more controlled environments. ByteDance has already signaled that its 'Maoxiang' app will continue to offer specialized agent services, suggesting that tech giants are moving these features away from their flagship general-purpose AI platforms to niche applications where monitoring can be more intensive. The shift reflects a broader strategy in Beijing to harness the productivity of AI while strictly curtailing its potential to disrupt traditional social structures and interpersonal relationships.