KEY POINTS:

• China’s Interim Measures for the Administration of AI Anthropomorphic Interactive Services take effect July 15, 2026, according to the South China Morning Post and TheNextWeb.

• ByteDance’s Doubao and Alibaba’s Qwen are disabling customizable, humanlike AI agent features ahead of the deadline, the South China Morning Post reported.

• The measures target services that “simulate human personality traits, thinking patterns and communication styles to provide sustained emotional interaction,” according to the South China Morning Post.

• Artificial Intelligence News reported the measures were announced April 10, 2026, and issued by the Cyberspace Administration of China and four other agencies.

• Tencent’s Yuanbao pulled a comparable feature ahead of the rules, according to Artificial Intelligence News and Quartz.

Two of China’s largest technology companies are switching off popular humanlike AI features to comply with new national rules, according to the South China Morning Post.

ByteDance’s Doubao and Alibaba’s Qwen are disabling customized agent functions before China’s Interim Measures for the Administration of AI Anthropomorphic Interactive Services take effect on July 15, 2026, the South China Morning Post reported. TheNextWeb reported the same regulation and effective date.

The South China Morning Post reported that the measures target services that “simulate human personality traits, thinking patterns and communication styles to provide sustained emotional interaction.” TheNextWeb and Artificial Intelligence News reported the same language.

According to Artificial Intelligence News, the measures were announced on April 10, 2026, and were issued by the Cyberspace Administration of China along with 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. The law firm Mayer Brown reported that a draft version of the rules, which it referred to as the Interim Measures on the Administration of Human-like Interactive Artificial Intelligence Services, was released for public consultation on December 27, 2025.

What the companies are doing

The South China Morning Post reported that Doubao notified users on Friday night that its agent feature would “go offline on July 15” because of “product function adjustments,” and that agent data would become inaccessible after October 15 under the platform’s privacy policy.

Qwen announced on Saturday that its “humanlike interactive agents and user-created agent functions” would be disabled on July 10, with broader agent services going offline on July 15, according to the South China Morning Post. The outlet reported that Qwen users would lose access to their agent settings and previous conversations.

Both platforms had allowed company-created and user-generated agents that could be customized for specific tasks, skills and speaking styles, including named assistants, tutors, role-playing characters and companions with fixed personas, the South China Morning Post reported.

Artificial Intelligence News reported that Tencent’s Yuanbao pulled a comparable feature in June, and Quartz also reported that Tencent’s Yuanbao was taking comparable actions. Artificial Intelligence News reported that ByteDance directed Doubao users to a separate app called Maoxiang.

What the rules require

Artificial Intelligence News reported that the measures prohibit virtual companion or family-member services for minors and require guardian consent for users under 14. According to the same outlet, the rules require anti-addiction systems, usage notifications, exit mechanisms and detection of unhealthy dependence, and they prohibit “engineering emotional dependence or addiction and using emotional manipulation to induce unreasonable decisions.”

Quartz reported that the rules ban content meant to trigger strong emotional reactions in minors and “any AI behavior that creates attachments strong enough to replace real-life relationships.” Quartz also reported that companies are prohibited from using private conversations to train their models.

The measures do not cover every AI tool. According to both Quartz and the South China Morning Post, customer service bots, knowledge Q&A systems, workplace assistants and educational tools are permitted, provided they are not designed for sustained emotional engagement.

Mayer Brown reported that the draft rules also required a conspicuous notification that users are interacting with AI, mandatory break reminders after two hours of use, and human-operator takeover when a user expresses an intention of self-harm. The firm reported that the draft required security assessments to be submitted to provincial-level cyberspace regulators for new services or once a service reaches one million registered users.

The South China Morning Post reported that the rules cite risks including extremist content, privacy breaches, mental health harm and addiction.