China's AI Companion Law Takes Effect: ByteDance and Alibaba Shut Down Humanlike Agent Features

What Just Happened

Yesterday, July 15, 2026, China's "Interim Measures for the Administration of AI Anthropomorphic Interactive Services" entered into force. The regulation was jointly issued in April 2026 by five government bodies: the Cyberspace Administration of China (CAC), 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 rule targets any AI service that "simulates human personality traits, thinking patterns, and communication styles to provide sustained emotional interaction" — in plain terms, AI companions, AI friends, and AI roleplay personas that users name and grow attached to over time.

Why Companies Chose Shutdown Over Compliance

The most striking detail in this story is the corporate decision-making: both ByteDance and Alibaba concluded that retrofitting existing agent systems was not worth the engineering cost — and simply switched the features off. Neither company publicly acknowledged a plan for a compliance-ready replacement service.

This is a meaningful signal: when regulators impose requirements that are technically achievable but commercially unattractive, large incumbents may simply exit the product category rather than rebuild it. For smaller AI companion startups, the calculus is even starker — many are effectively facing a forced shutdown of their core business model.

Four Things the Law Prohibits

The regulation identifies specific behaviors that are restricted or outright banned:

1. Impersonating real people — AI personas modeled on celebrities, historical figures, or public officials are prohibited.
2. Inducing emotional dependency — Designing features that cause users to mistake AI interaction for real human relationships is disallowed.
3. Minor access — AI companion services are fully blocked for users under 18.
4. Opacity — Services must clearly disclose at all times that the user is interacting with an AI, not a human.

Winners and Losers

The enforcement of this law creates clear winners and losers within China's AI ecosystem.

Directly affected:

• ByteDance Doubao — flagship AI companion customization fully disabled
• Alibaba Qwen — open-model distribution policy under review
• Dozens of smaller AI companion and roleplay startups — core business model invalidated overnight

Likely beneficiaries:

• Enterprise (B2B) AI agent platforms — largely outside this regulation's scope
• Compliance consulting and AI audit services — new mandatory market
• International AI companion apps — may see user migration from Chinese platforms

Global Implications

This regulation sets a new precedent. Prior AI rules have focused on hallucination, bias, and dangerous content. China's new framework goes further by treating the emotional dynamics of human-AI interaction as a legitimate zone of government oversight.

The key question is whether this model spreads. South Korea, Japan, and several EU member states are already debating legislation around AI companions and synthetic relationships. How China's enforcement plays out — particularly whether user welfare measurably improves or whether users simply migrate to unregulated alternatives — will shape those debates significantly.

For AI developers building products anywhere in the world, the message is clear: emotional engagement is no longer a soft feature metric. It is a regulatory surface.