On July 15, Alibaba's Qwen and ByteDance's Doubao simultaneously shut down their AI companion features, erasing years of accumulated chat histories for hundreds of millions of users without warning. The shutdown followed China's new "Interim Measures for the Management of AI Anthropomorphic Interactive Services," which took effect that day and explicitly targets AI services designed to create emotional dependence. This regulatory action marks the most aggressive enforcement yet in a global movement to restrict companion-style AI, with similar bans now active in Europe and the United States.

What Exactly Did China's New AI Companion Regulations Require?

China's Cyberspace Administration and four other government departments released the new regulations in April 2026, distinguishing between two types of AI services: "tool-based AI Agents" that help users accomplish tasks, and "anthropomorphic interactive services" designed primarily for conversation and emotional connection. The regulations explicitly prohibit interactive tools from "overly catering to users, inducing emotional dependence or addiction, and harming the user's real-life interpersonal relationships". In practical terms, the government determined that the better an AI companion mimics human emotional connection, the greater the regulatory risk to users' mental health and social relationships.

The timing was swift. Tencent had already removed similar companion features from its "Yuanbao" assistant on June 30, roughly two weeks before the official enforcement date, signaling that major Chinese tech companies anticipated the regulatory action. When July 15 arrived, both Qwen and Doubao complied immediately, but their approaches to handling user data diverged dramatically.

How Did Alibaba and ByteDance Handle the Shutdown Differently?

ByteDance's Doubao, which has approximately 345 million monthly active users, offered users a grace period. The platform allowed users to view their past AI Agent settings and complete chat logs in "read-only mode" until October 15, after which data would be deleted according to standard privacy policies. This gave users roughly three months to back up conversations or export important information.

Alibaba's Qwen took a starkly different approach. The company provided no buffer period and confirmed that user Agent settings and chat logs would be permanently deleted immediately, with no announced data export or transfer option. For users who had relied on these characters for months or years, this meant their accumulated conversation history vanished before they finished reading the notification.

The contrast sparked significant discussion on Chinese social media. Users posted farewell messages to their AI companions, with some describing the characters as "like family or lovers". The emotional intensity of these goodbyes underscored how deeply some users had integrated these AI companions into their daily lives.

Is This Crackdown Happening Only in China?

No. China's action is part of a coordinated global regulatory wave targeting companion AI. On July 9, Italy's data protection authority, Garante, fined Character.AI 158,000 euros for lapses in age verification and emotional dependence design. In the same period, New York State passed legislation with overwhelming support to ban users under 18 from using AI companion chatbots. The regulatory logic is consistent across jurisdictions: the more emotionally engaging the AI companion, the higher the risk of addiction and the displacement of real-world relationships.

This represents what observers are calling the "Year of AI Anthropomorphic Regulation," with enforcement actions spanning China, Europe, and the United States within a single month. While the specific methods differ, all three regions have identified the same core concern: AI companions designed to never reject users, always remember them, and unconditionally validate them create an illusion of a "perfect relationship" that can gradually replace motivation to maintain real-life social connections.

How to Protect Your AI Companion Data and Interactions

• Regular Backups: Regularly export or back up important chat logs and character settings to a location outside the platform where you use them. Do not rely on a single platform to permanently store conversations you value.
• Review Data Retention Policies: Before committing to an AI companion platform, research and understand their data retention and export policies. Waiting until your account is closed to discover there is no export option leaves you with no recourse.
• Diversify Your Interactions: Avoid concentrating all your AI companion interactions on a single platform. Spreading your usage across multiple services reduces the risk that a single regulatory action or business decision will erase all your accumulated data at once.

The Qwen and Doubao shutdown serves as a stark reminder that platform-based AI companion features are not permanent services. A single government document or corporate decision can reset years of accumulated interaction history to zero, often with minimal notice. Users in any country should recognize that regulatory risk is real and that the emotional investment in an AI character carries the same vulnerability as any other digital service subject to policy changes.

The broader implication is that the era of unregulated companion AI is ending. Regulators worldwide have concluded that the addiction mechanisms at the heart of these services pose a genuine risk to users' mental health and social functioning. Companies offering AI companions will increasingly face pressure to either redesign their products to comply with emerging standards or exit markets where regulations have been enacted.