Doubao and Tongyi Qianwen Roll Back AI Personification Tools Amid Tightening Regulatory Controls

On July 15, the Interim Measures for the Management of Human-like Interactive Services of Artificial Intelligence will officially take effect. Two of China’s leading AI platforms have already begun responding to these new requirements. ByteDance’s “Doubao” and Alibaba Group’s “Tongyi Qianwen” are planning to remove their customizable human-like agent features, becoming the first major platforms to adjust their offerings ahead of the regulation’s implementation.

Doubao sent a notification to its users on Friday evening announcing that it would stop offering its agent function starting July 15, citing “product function adjustments” as the reason. The message also stated that any data related to this feature would be handled in accordance with the company’s privacy policy beginning October 15, after which users would no longer be able to view or retrieve such information within the app. Tongyi Qianwen is moving faster; it will disable its “human-like interactive agents and user-created agent functions” by July 10, while all of its “Tongyi agent functions and services” will cease operations entirely on July 15. Once that date arrives, users will no longer have access to their existing agent settings or previous conversation records.

The new regulations focus on preventing “continuous emotional interaction,” prompting these three major tech companies to scale back their related features

Previously, both platforms offered agent pools developed jointly by the company and its users, allowing them to set fixed personalities and communication styles for AI assistants. These assistants could serve as named helpers, learning tutors, role-playing characters, or emotional companions. The timing of these functional changes aligns closely with the rollout of the regulatory measures. The Interim Measures for the Management of Human-like Interactive Services of Artificial Intelligence, published in April, define as regulated services those that “simulate human personality traits, thinking patterns, and communication methods while providing continuous emotional interaction.” However, customer service robots, knowledge question-answering tools, workplace intelligent assistants, and educational or research tools are excluded from this scope. The regulations highlight potential risks such as the spread of extreme ideas, privacy breaches, harm to users’ physical and mental health, and the development of dependency or addictive behaviors.

It is worth noting that Tencent had already removed similar agent functions from its AI assistant “Yuanbao” back in June. To date, China’s three major technology giants have shown a consistent trend of reducing their human-like agent capabilities. Based on these regulatory signals, China appears intent on integrating AI agents into future productive infrastructure while imposing stricter controls on human-like companion agents that may form emotional or quasi-social connections. Between May and June of this year, regulators also issued guidance aimed at promoting the orderly development of AI agents, along with national standards for interoperability, to establish unified systems covering identification, authorization, connectivity, and traceability.

User reactions: Loss of emotional support comes suddenly

The introduction of these policies has triggered some negative feedback from users. One Weibo user associated with the Doubao official account wrote, “Why are we removing the agent? It has always been our emotional support.” The user also expressed concern that long-term chat histories and the emotional bonds formed through interactions are difficult to export or transfer using current methods, and the lack of a smooth data migration process makes this change feel particularly abrupt. When AI companions become an essential source of comfort for some users, there is clearly a need for more flexibility and transition time when balancing regulatory requirements with product development.