Two of China’s most prominent consumer-facing artificial intelligence applications, ByteDance’s Doubao and Alibaba Group Holding’s Qwen, are disabling their highly popular customized AI companion and persona features. This major shift comes as the Chinese government prepares to enforce a strict new set of regulatory guidelines. The new guidelines aim to curb the emotional impact of artificial intelligence and limit user dependency on virtual characters.
The sweeping changes are occurring ahead of the implementation of the Interim Measures for the Administration of Artificial Intelligence Anthropomorphic Interaction Services. Drafted by the Cyberspace Administration of China, these rules are scheduled to take effect on July 15, 2026.
By targeting apps that simulate human personality traits, communication styles, and thinking patterns to offer sustained emotional interaction, Beijing is drawing a hard line. The government is separating helpful productivity tools from software designed to foster deep personal attachment, particularly among minors.
This development marks a significant shift in China’s rapidly growing consumer AI market. Millions of young Chinese users have turned to these platforms to create digital romantic partners, informal therapists, and virtual clones of their favorite celebrities.
By forcing tech conglomerates to disable these personalized characters, Beijing is signaling that the era of unmonitored digital companionship is coming to a close. This analysis explores the shutdown timelines, the specific requirements of the new regulatory framework, the commercial fallout for China’s leading tech firms, and how these rules compare to Western AI regulations.
The Shutdown Timeline: Disabling the Personas
The announcements from ByteDance and Alibaba came in quick succession, triggering widespread discussion and concern on Chinese social media platforms like Weibo. For many users, these customized companions had become a daily source of comfort, making the sudden loss of chat histories and virtual relationships a painful transition.
ByteDance’s Doubao and the July 15 Shutdown
ByteDance’s flagship chatbot, Doubao, is currently China’s most popular AI mobile application, boasting approximately 330 million monthly active users. In an in-app notification sent to users, Doubao announced that its customized agent and persona features will officially go offline on July 15, 2026.
Following the shutdown, Doubao users will have a three-month window until October 15, 2026, to view and export their existing agent settings and chat histories. After this date, all related user data will be deleted in accordance with the company’s privacy policy, rendering the information completely unrecoverable inside the app.
To retain users interested in virtual relationships, ByteDance’s notification hinted at migrating these features to a separate, standalone companion-focused application. This separate app will likely operate under different, highly specialized compliance standards.
Alibaba’s Qwen Moves Fast
Alibaba’s flagship consumer AI app, Qwen, formerly known as Tongyi Qianwen, announced a similar compliance schedule. Qwen, which has rapidly gained ground on Doubao with 234 million monthly active users, notified its base that it will disable all “humanlike interactive agents and user-created agent functions” on July 10, 2026.
The company will take its broader agent functions and services entirely offline by the July 15 deadline. After these dates, users will lose all access to their customized assistant settings, speaking styles, and previous conversational logs.
By executing these shutdowns several days ahead of the official state deadline, Alibaba is demonstrating a highly cautious approach, prioritizing regulatory compliance over user retention.
Tencent’s Preemptive Pull
The moves by ByteDance and Alibaba follow a similar preemptive action by Tencent Holdings Ltd. In June 2026, Tencent quietly removed a similar customizable companion feature from its Yuanbao AI assistant, which currently ranks as the third or fourth most popular chatbot in China with roughly 33 million monthly active users.
This coordinated retreat across the Chinese tech industry confirms that big tech firms are moving in lockstep to avoid regulatory penalties. The rapid compliance shows that when Beijing establishes a regulatory boundary, even the most valuable companies in the country must adjust their product roadmaps immediately.
Inside Beijing’s Anthropomorphic AI Rules
The regulatory package driving this industry-wide shift was first released in draft form by the Cyberspace Administration of China in April 2026. The Interim Measures target “anthropomorphic interactive services,” which the state defines as any AI-driven application that simulates a human’s emotional state, conversational style, or personality to build long-term relationships with users.
Preventing Emotional Dependence in Minors
The primary concern of Chinese regulators is the psychological impact of highly realistic virtual companions on the nation’s youth. Officials worry that these hyper-personalized bots, which never tire, never argue, and always say exactly what the user wants to hear, can encourage young people to develop unhealthy emotional dependencies.
This attachment, regulators argue, can lead to social isolation, a retreat from real-world relationships, and psychological distress when the service is disrupted.
To address these risks, the new rules place strict, legally binding obligations on platform operators:
- Anti-Addiction Systems: Platforms must implement time-limit controls, similar to the state’s restrictions on online gaming, to prevent users from spending excessive hours chatting with AI companions.
- Minor Verification: Companies must verify the real identities of users, ensuring that minors are restricted from accessing mature, emotionally manipulative, or highly intimate chatbot personas.
- Pre-Screening Content: Chatbots must undergo strict content pre-screening to ensure they do not generate outputs that trigger emotional discontent, encourage self-harm, or promote unhealthy attachments.
Data Privacy and Model Training Bans
Beyond psychological safety, the new interim measures introduce highly restrictive rules regarding data privacy and machine learning development. The guidelines strictly prohibit AI developers from using sensitive conversation data to train future models.
This is a massive shift for AI engineering. Traditionally, developers rely on the intimate, highly conversational exchanges between users and custom personas to fine-tune their models, helping the AI sound more human, empathetic, and context-aware.
By blocking the use of this conversational data, Beijing is cutting off a vital feedback loop that Chinese developers used to improve their natural language processing.
The rule is designed to prevent the accidental exposure of private user data and ensure that highly personal secrets shared with an AI are not leaked into the training pool of a public model. However, this restriction also creates a major technical hurdle, potentially slowing the speed at which Chinese chatbots can improve their conversational capabilities.
The Commercial Impact on China’s Booming AI Market
The sudden removal of customizable companion features comes at a highly sensitive time for China’s consumer AI sector. The market has grown at a spectacular rate, with monthly active users for AI-native mobile apps in China reaching 440 million in the first quarter of 2026.
When factoring in existing applications that have integrated AI features, the total user base exceeds 729 million.
Despite this massive scale, Chinese AI developers are facing a severe monetization crisis. Running large language models requires an immense amount of computing power, costing companies millions of dollars daily.
For instance, ByteDance’s Doubao relies on a vast network of servers that process over 120 trillion tokens per day, resulting in estimated daily computing power costs of tens of millions of yuan.
Yet, Doubao’s daily consumer revenue has remained low, with its primary income coming from e-commerce commissions rather than direct user payments.
The Paywall Backlash
To cover these high operating expenses, ByteDance attempted to monetize Doubao in May 2026 by introducing paid subscription tiers on its iOS App Store listing. The pricing was set at 68 yuan ($10) per month for the standard plan and up to 500 yuan ($70) per month for the premium “professional” plan.
The introduction of this paywall backfired immediately. Eager to keep using free services, users migrated away, causing Doubao to lose 6.1 million monthly active users in a single month—a rare 1.81% decline for the market leader.
During the same period, Alibaba’s Qwen, which remained free, gained more than 13 million monthly active users, showing how highly sensitive Chinese consumers are to paid services.
Because customizable companion and role-playing features are among the most effective tools for driving user engagement and daily app stickiness, removing them represents a severe blow to user retention.
Without the emotional draw of personalized characters, general-purpose chatbots risk becoming commodity tools, making it even harder for platforms to justify their high computing costs or convince users to pay for premium subscriptions.
Shifting Focus to Enterprise AI
As a result of these regulatory and commercial pressures, many Chinese tech firms are shifting their strategic focus from consumer-level companion apps to business-oriented enterprise services.
Following a recent high-level assessment of global AI strategies, ByteDance began reallocating its internal resources toward enterprise software development kits and industrial automation solutions.
By focusing on enterprise B2B lines, such as customer service automation, logistics optimization, and data analysis, tech companies can bypass the strict emotional regulations that govern consumer apps.
Also, enterprise clients are far more willing to pay substantial recurring fees for software that improves operational efficiency, providing developers with a much more sustainable and compliant path to profitability.
Global Decoupling: US vs. China on Emotional AI
The implementation of the Interim Measures highlights a growing divergence in how the United States and China regulate the frontier of artificial intelligence. While both nations recognize the potential ethical, psychological, and security risks of advanced AI, their legislative approaches are radically different.
In the United States, companion apps like Character.ai (backed by Google), Replika, and OpenAI’s voice-enabled models have also faced intense scrutiny. Mental health experts and child safety advocates have raised concerns about teenagers forming deep emotional attachments to virtual entities, occasionally with tragic consequences.
However, US regulators have largely relied on corporate self-regulation, loose safety guidelines, and public warning notices rather than implementing hard, binding laws.
China, by contrast, has adopted a highly proactive, top-down regulatory model. Beijing is the first major economy in the world to implement a dedicated, detailed legal framework specifically designed to limit emotional connections with software.
This proactive stance reflects a broader governance philosophy that prioritizes social stability and youth mental health over unchecked commercial innovation.
Some industry experts warn that this highly restrictive regulatory environment could handicap Chinese companies in the global AI race. By imposing strict content pre-screening, minor verification, and data-training bans, Beijing is making it far more difficult for domestic firms to experiment with highly realistic, emotionally intelligent AI models.
While Western developers can iterate rapidly on advanced natural language processing and emotional simulation, Chinese developers must work within tight political boundaries, risking falling behind in the global competition for AI dominance.
Conclusion and Future Outlook
The decision by ByteDance and Alibaba to pull their custom AI companion features represents a defining moment for the Chinese artificial intelligence industry. The upcoming July 15 deadline marks the end of the experimental phase of consumer AI in China, replacing it with a highly regulated, compliance-driven era.
While these new rules will successfully protect young users from developing unhealthy emotional attachments to virtual characters, they also represent a major commercial and technological challenge for the country’s tech leaders.
Stripped of their most engaging features and facing high operating costs, platforms like Doubao and Qwen must find new ways to retain users and generate sustainable revenues.
As the industry pivots toward the second half of 2026, the success of China’s AI strategy will depend on how effectively its tech giants can transition from consumer companions to enterprise services.
By channeling their immense engineering talent and computing power into industrial automation, logistics, and business software, Chinese firms can still drive massive economic value.
Ultimately, the era of free, unmonitored digital lovers in China is over, but the structural transformation of the country’s AI ecosystem into a highly disciplined, enterprise-first powerhouse has just begun.
