July 5, 2026, (Inside AI) — In a sudden regulatory pivot, China’s tech titans ByteDance and Alibaba are pulling the plug on their AI agent platforms. ByteDance’s Doubao app and Alibaba’s Qwen service will disable user-created agents and humanlike interactive functions this month, directly aligning with new national rules on anthropomorphic AI.
The shutdowns, announced just days apart, target features that let users craft chatbots with persistent personas, emotional tones, and specific skills. Doubao will disable its agent creation and humanlike interaction features on July 10. Qwen’s similar functions go dark the same day, with all remaining agent services offline by July 15.
This isn’t a technical glitch. It’s a compliance sprint. The Interim Measures for the Administration of Artificial Intelligence Anthropomorphic Interaction Services, issued in April, come into force on July 15. The rules target AI that “simulate human personality traits, thinking patterns and communication styles to provide sustained emotional interaction.”
Both apps hosted sprawling agent marketplaces. Users could transform a generic chatbot into a named tutor, a role-playing companion, or a digital friend with a fixed backstory. Qwen’s notice confirmed that after shutdown, users lose access to agent settings and conversation histories entirely.
The measures explicitly exempt customer service bots, knowledge Q&A, workplace assistants, and educational tools—unless they foster ongoing emotional bonds. The government’s concerns are stark: extremist ideas, privacy breaches, psychological harm, and addiction.
China’s Regulatory Calculus on Emotional AI
This crackdown didn’t emerge in a vacuum. For years, China has tightened control over algorithmic recommendation systems and deep synthesis technologies. Now, the focus shifts to relational AI—systems designed to mimic human connection. The new rules define anthropomorphic interaction as sustained emotional engagement, a line that many consumer AI agents cross by design.
ByteDance and Alibaba aren’t alone. Other Chinese firms with similar agent platforms will likely follow. The measures give regulators broad authority to audit training data, enforce transparency, and mandate user protection safeguards. Non-compliance risks fines, service suspension, or worse.
Industry observers note a parallel with earlier curbs on livestreaming and social media algorithms. The state views uncontrolled emotional manipulation as a societal risk. An AI companion that learns a user’s vulnerabilities could exploit them, whether for profit or ideological influence.
Global context sharpens the contrast. While Western companies race to deploy empathetic AI for therapy, eldercare, and entertainment, China is slamming the brakes. Character.AI and Replika face no such federal shutdowns in the U.S., though they grapple with their own content moderation crises.
What Gets Lost When Agents Go Dark
Behind the policy language, real use cases vanish. Doubao and Qwen agents served as language tutors, creative writing partners, and even informal mental health sounding boards. Some were built by third-party developers, forming nascent ecosystems. Their removal resets expectations for consumer AI in China.
Users who invested time crafting personalized agents will lose that work. The psychological impact on those who formed attachments—however one-sided—remains unstudied. Qwen’s blunt notice offered no data export or migration path.
The rules carve out exceptions, but the boundaries are fuzzy. A workplace assistant that learns a user’s communication style could inadvertently become “anthropomorphic.” Companies must now audit every interaction for emotional depth, a technically daunting task.
This also reshapes the competitive landscape. Startups betting on emotional AI as a differentiator must pivot or perish. International firms eyeing China’s market now face a clear signal: keep your bots transactional, or stay out.
ByteDance and Alibaba will likely retool their platforms for compliant, task-oriented agents. But the era of openly playful, emotionally sticky AI in China ends on July 15. The question now is whether other jurisdictions watch and learn—or dismiss this as a uniquely Chinese experiment in AI governance.
