Regulation arrived sooner than most people expected, and two of China’s biggest artificial intelligence companies are already making moves to adapt.
ByteDance and Alibaba announced they will shut down so-called custom agents from their most popular AI apps, Doubao and Qwen.
The reason is pretty straightforward: new Chinese government rules take effect on July 15, and they significantly change the game for this type of feature. 📅
These agents were, in practice, virtual assistants with their own personality that users could create and customize however they wanted, from a patient tutor to a conversation companion with a defined name, tone, and style.
And that is exactly the kind of interaction the new measures want to regulate more carefully.
What is behind this decision goes beyond a simple product update. It involves a very clear government vision for how artificial intelligence should interact with people.
The deadlines users need to know
Doubao, by ByteDance, notified its users in a statement published Friday evening that the agent feature would go offline on July 15, citing product function adjustments. According to the company, after October 15, data related to those agents would be handled in accordance with the company’s privacy policy and would no longer be available for viewing or recovery within the app.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
Qwen, by Alibaba, followed the same path and issued a similar notice Saturday morning. The company said its interactive agents with human-like appearances and user-created agent functions would be deactivated on July 10, while Qwen’s broader agent functions and services would go offline on July 15. After the shutdown, users would no longer be able to access agent settings or previous conversations.
Worth noting that Alibaba itself owns the South China Morning Post, the outlet that originally reported these changes. Neither ByteDance nor Alibaba immediately responded to requests for comment made on Sunday. ⏰
What these agents were and why people liked them so much
Before understanding the impact of regulation, it is worth understanding what is actually being shut down. The custom agents on Doubao and Qwen were not simply generic chatbots. Both apps offered a lineup of agents, built by the companies themselves and by users, that could be customized for specific tasks, skills, and speaking styles. Users could turn a general-purpose chatbot into an assistant with its own name, a tutor, a roleplay character, or even a companion with a fixed persona and tone.
This represented a level of customization that few artificial intelligence products had achieved in such an accessible way for everyday users. A person could create an agent with a name, communication style, area of expertise, and even a specific personality, all without needing to know how to code or understand anything about machine learning. It was literally a tailor-made assistant that responded in a way that felt natural and personal.
This created an extremely engaged user base, especially among young adults who used these agents for purposes ranging from studying to entertainment and informal emotional support. Many people reported daily conversations with these assistants, using them to organize their routines, practice languages, discuss ideas, and even process everyday situations. The bond that formed between users and their custom agents was something developers clearly did not fully anticipate, and it is precisely that bond that raised red flags for Chinese authorities.
From a user experience standpoint, the feature was an almost perfect example of how artificial intelligence can become part of daily life in a smooth and organic way. Personalization reduced friction in interaction, made responses more relevant, and created a sense of continuity that traditional chatbots rarely achieve. The problem, according to the Chinese government, is that this fluidity also opens the door to risks that need to be managed with clear and structured rules.
What the new rules require and why it matters
The timing of this change coincides with the implementation of the Interim Measures for the Administration of Anthropomorphic Interaction Services of Artificial Intelligence, which take effect on July 15. Published in April, the rules cover AI services that simulate human personality traits, thought patterns, and communication styles to deliver continuous emotional interaction.
It is important to understand that the rules do not apply to everything. They exclude customer service bots, knowledge-based question-and-answer systems, work assistants, and tools focused on education and scientific research, as long as those applications do not involve continuous emotional interaction. In other words, the focus is squarely on agents that create emotional or quasi-social bonds with users.
The measures cited specific risks that motivated the creation of the rules, including the spread of extremist ideas, privacy breaches, harm to physical and mental health, and the risk of dependency or addiction. The central concern among authorities is that agents with a personality built by the user can create emotional dependencies, distort the perception of reality, or even be used in ways that fall outside the control of developers and platforms.
For ByteDance and Alibaba, adapting goes beyond simply turning off a feature. They need to rethink how their artificial intelligence products will remain relevant to users without the deep personalization that was one of the biggest competitive differentiators for these platforms. This involves product decisions, changes in system architecture, and potentially the development of new interaction formats that are both appealing to users and within the boundaries set by the new rules. 🔧
Tencent already made the first move
ByteDance and Alibaba were not the first to touch this kind of feature. In June, Tencent Holdings removed a similar function from Yuanbao, its consumer-facing AI assistant app. This shows the movement is part of a broader trend among major Chinese tech companies getting ahead of regulatory requirements.
Pan Helin, a member of the expert committee of the Ministry of Industry and Information Technology, commented that using agents requires a certain level of understanding from users. According to him, current agents are still not mature enough. This assessment reinforces the perception that the technology moved fast, but the collective understanding of its limits and risks is still catching up.
The bigger picture: China and the race to regulate AI
Chinese regulators are paying increasing attention to AI agents, especially because the technology has moved beyond chatbots and reached systems capable of maintaining memory, planning actions, triggering tools, and executing tasks. This completely changes the level of complexity and responsibility involved in these products.
In May, regulators issued guidelines on the managed development of AI agents, recognizing them as an important form of artificial intelligence products and services while also calling for safety measures, controls, and practical application. Pan highlighted that the policy emphasized three priorities: safety, practical use, and standardization.
Then in June, China released a series of national standards for AI agent interconnection, covering architecture, identity codes, identity management, agent description, discovery, interaction, and tool usage. This signaled Beijing’s determination to make agents identifiable, authorized, connected, and traceable. Taken together, these measures suggest China wants to encourage AI agents as part of its productivity infrastructure while tightening controls on companion agents that could form emotional or quasi-social relationships with users.
For companies like ByteDance and Alibaba, operating within this regulatory environment is an ongoing reality. Unlike markets where regulation comes after mass adoption, China has a much more dynamic relationship between innovation and oversight. Companies develop, the government watches, and when boundaries are crossed or risks become visible, the rules arrive. This cycle accelerates regulatory maturity but also demands a capacity for rapid adaptation that few players can sustain without significant operational impact. 🌐
User reactions to the shutdown
The abrupt removal of agents sparked reactions from some users, who expressed frustration on social media. In a Weibo post, a user going by the name Tuxiaoxiao tagged the official Doubao account and questioned why the agents were being pulled, saying they had been an emotional support for a long time.
The same user complained about the amount of conversation records and all the sentiment built over so much time, also pointing out that there was no simple way to export or transfer that data. This kind of reaction illustrates well the bond many people formed with these assistants, and it also exposes a sensitive point: when a feature this personal is shut down, the question of data portability and preservation becomes central.
What to expect going forward
With the shutdown of custom agents, the artificial intelligence market in China enters a new phase. Companies will need to find creative ways to maintain user engagement without relying on the deep personalization that was at the core of the experience. This could open up space for innovation on other fronts, such as collective agents, context-based features, and experiences more oriented toward tasks than virtual relationships. Users who were used to an agent that seemed to know them well will probably miss it, but adaptation tends to happen, especially if products deliver value in other ways.
Another important point is the effect this movement could have globally. When two of the world’s largest artificial intelligence companies are forced to reconfigure their products because of regulation, it sends a clear signal to the rest of the industry. Companies in other countries developing similar custom agent features are certainly watching what is happening in China, not necessarily to copy the regulatory model, but to understand which sensitive areas tend to attract government attention around the world.
On the horizon, what is taking shape is an artificial intelligence ecosystem increasingly shaped by rules that vary from country to country, creating complex challenges for global platforms and opportunities for those who know how to navigate these different environments well. ByteDance and Alibaba are now on the front lines of this process, and how they reinvent themselves over the coming months will be an important chapter in the story of AI. 🤖
