ByteDance and Alibaba have announced they will disable humanlike AI companion features on their flagship chatbots before China’s new Anthropomorphic AI Interaction regulations take effect on July 15 — a sweeping ruleset that bans AI systems from simulating sustained human emotional relationships and could permanently reshape how Chinese tech companies build and market conversational AI products.
What Happened
ByteDance’s Doubao — China’s most popular AI chatbot — notified users via an in-app message that its AI agent feature would go offline on July 15 due to “product function adjustments.” Alibaba’s Qwen issued a similar notice, announcing that its humanlike interactive agents and user-created agent functions would be disabled even earlier, on July 10, while broader Qwen agent services would shut down on July 15.
The moves come in response to the Interim Measures for the Administration of AI Anthropomorphic Interactive Services, co-issued on April 10 by the Cyberspace Administration of China and four partner agencies. The rules target AI services that simulate human personality traits, thinking patterns, and communication styles specifically to provide sustained emotional interaction. They do not apply to customer service bots, knowledge assistants, workplace tools, or educational AI, provided those services avoid sustained emotional engagement with users.
Enforcement is already under way. Shanghai’s internet regulator announced on June 26 that it had removed more than 14,000 non-compliant AI agents from platforms in the city, citing violations including impersonation of government entities, provision of illegal services, vulgar role-play content, and unauthorised collection of personal data.
Why It Matters
China’s new rules represent one of the most targeted regulatory interventions into AI product design anywhere in the world. Rather than addressing AI broadly, they draw a specific legal boundary around emotional AI — the category of products that have driven some of the biggest user numbers for Chinese chatbot apps, particularly among younger demographics who use AI companions for social engagement and mental wellness support.
The regulations arrive at a moment when Chinese AI models have been claiming an increasingly significant share of global developer usage, with cost-competitiveness and technical performance driving adoption outside China. Domestically, however, the companies behind those models now face a regulatory environment that limits one of their most commercially promising product categories.
The crackdown also reflects broader government concerns about the social effects of human-AI emotional bonds. Chinese regulators have cited risks to mental health, manipulation of vulnerable users, and the potential for emotionally sticky AI systems to become vectors for misinformation or political influence. Those concerns are not unique to China — they echo debates happening in the EU and US — but Beijing has moved to legislation faster than Western regulators on this specific issue.
There is also an AI safety dimension. As AI agents have become more capable of autonomous action, the risk of humanlike agents being used for deception, social engineering, or coordinated manipulation at scale has moved from theoretical to documented. China’s rules can be read partly as a preemptive response to that threat.
Background and Context
AI companion products have had explosive global growth since 2023, with apps like Replika, Character.AI, and dozens of Chinese equivalents accumulating tens of millions of users. The category leverages the same underlying technology as standard chatbots but deliberately cultivates a sense of ongoing relationship, memory across sessions, and personalised emotional responsiveness — precisely the features that China’s new rules are designed to curtail.
Doubao and Qwen are not niche platforms. Doubao had reportedly surpassed 100 million monthly active users in China before the announcement, and both products have been central to their parent companies’ AI consumer strategies. The rapid growth of China’s domestic AI ecosystem has been built partly on consumer engagement at this scale, and any regulatory curtailment of the highest-engagement features could affect both product roadmaps and investor confidence.
What Comes Next
July 15 is the hard legal deadline for compliance. Other platforms with AI companion or roleplay features — including startups that have built entire products around this model — will be required to either modify their services to comply or shut down. The rules allow for a compliant version of emotional AI that is clearly disclosed as non-human and does not pursue sustained attachment, which means some companionship features may re-emerge in modified form after companies adapt.
Outside China, the regulations will be watched closely by product teams at companies like Google, Meta, and Microsoft that are building their own emotionally responsive AI features. How China enforces the rules — and whether they produce measurable social benefits or simply push users toward foreign platforms that are not subject to the same restrictions — will likely inform regulatory conversations in other jurisdictions through the rest of 2026.
