ByteDance’s Doubao, China’s most popular AI chatbot, is disabling a core feature today. Users can no longer customize their own AI personas—a feature that let people build unique AI companions tailored to their preferences.

The shutdown comes as Beijing’s new rules on “anthropomorphic” AI services took effect on July 15. The rules target AI systems designed to simulate human personalities and emotional relationships. Alibaba’s Qwen and Tencent’s Yuanbao are making similar cuts.

What the New Rules Ban

The Interim Measures for the Administration of Artificial Intelligence Anthropomorphic Interaction Services—issued in April, effective today—prohibit AI behavior that creates attachments strong enough to replace real-life relationships. The rules also ban content meant to trigger strong emotional reactions in minors.

Companies can no longer use private user conversations to train their models. The rules are sweeping enough that they’ve forced China’s biggest AI platforms to strip features almost overnight.

Broader Implications for China’s AI Industry

This is just the visible enforcement. Behind the scenes, Chinese authorities are holding talks with Alibaba, ByteDance, and Z.ai about whether to restrict foreign access to their most advanced models entirely, Reuters reported. Nothing is decided yet, but officials are sketching options—including a blanket ban on public release or a limit to domestic use only.

The message is clear: Beijing wants control over how AI interacts with people. Customizable AI companions blur the line between tool and relationship. China’s regulators want that line back.

For global AI companies watching China, today marks a shift from monitoring regulation to enforcing it. How the policy lands will matter for how other countries shape their own rules.