What If ByteDance’s Doubao AI Smartphone Belonged to Tencent?

By Ma Jinnan

The wave of AI-powered smartphones has yet to truly sweep the market, but the first wave of “ecosystem barriers” has already emerged.

Recently, ByteDance partnered with ZTE to try to get ahead of the competition by launching an engineering prototype of the Nubia M153 equipped with the Doubao AI assistant. However, they unexpectedly ran into a “fortress wall” of reality. This prototype—designed to deeply integrate AI capabilities at the system level—triggered risk alerts from several national-level apps such as WeChat due to its system-level permission invocation strategy. As a result, users experienced abnormal logouts and even restrictions on account access. Clearly, this is an awkward start for the generalization of AI capabilities in mobile phones and casts a shadow over the future of Doubao AI-powered smartphones.

Is this a conflict about technological paths?

No.

Is this a clash over privacy protection strategies?

No.

This is the inevitable eruption of a battle for dominance at the gateway to the AI era.

In the mobile internet age, where tech giants have staked their territories and ecosystems are highly fragmented, how can an “outsider” like an AI assistant gain the “system’s” trust? If a tech giant with even stronger ecosystem control, like Tencent, attempted to launch an AI smartphone, would the situation play out differently?

Let’s first take a look at why the Doubao AI phone was “blocked.”

From a technical perspective, the AI assistant did indeed “overstep its bounds”—it crossed the security red lines of many app developers.

The so-called “system-level permission invocation” is one of the highest-level permissions in a smartphone’s operating system, allowing an app to inject simulated user input events—such as taps, swipes, and key presses—into the system. For an AI assistant, this means it has an “invisible hand” and can operate any app just like a real person. It also means the AI assistant can deeply intervene in, or even control, the operations of other apps—such as reading screen information or simulating user taps, and more.

The essence of this technical capability is that the phone manufacturer uses its own system private key to sign and set the Doubao AI Assistant as an integral part of the operating system.

This means that Doubao AI is no longer just an assistant, but rather becomes the “hand of God.”

For social and financial applications represented by WeChat, this is tantamount to installing a “third-party camera” in their own bedroom. To protect users’ vast amounts of sensitive data and account security, such apps have extremely sensitive risk control mechanisms—any unofficial or abnormal system-level call will be regarded as a potential attack, triggering “self-protection” bans as an instinctive response.

On the surface, the dilemma faced by Doubao AI phones seems like a tug-of-war over technical permissions. But if we look deeper, we'll find that the real issue is a conflict of ecological positions, as well as a lack of trust and an ongoing contest of interests between the new force of AI attempting to break down existing “walled gardens” and the giants of the ecosystem.

ByteDance and Tencent have long been direct competitors in the battles for content, traffic, and user attention. Letting an AI assistant developed by a rival “take charge” across the board and deeply intervene in their core application ecosystem hardly makes sense from a business perspective. Moreover, as a new AI product, Doubao’s capabilities in data security and privacy protection have yet to stand the test of time in the market, giving leading national apps no justification to entrust the trust of hundreds of millions of users to an “outsider.”

So, what is ByteDance’s real intention behind launching the Doubao AI Assistant?

Traditionally, users need to open multiple apps and switch among them to satisfy their social, shopping, and information needs. With Doubao Assistant, complex workflows across various apps can be completed with a single voice command—essentially “short-circuiting” the traditional traffic gateways of these apps.

This is about seizing supreme control over users’ attention and the distribution of traffic.

Imagine a future where users are accustomed to saying things like "Send my boss a New Year greeting on WeChat," "Order takeout for me," "Help me compare prices and shop," or "Book my flight" directly to their AI assistant. In this scenario, the entire process that used to require opening an app, browsing, and placing an order gets compressed. The value of traffic gateways on platforms like WeChat, Taobao, Meituan, JD.com, Didi, and Ctrip would be significantly diminished.

Behind the Doubao AI Assistant is ByteDance—a player that has repeatedly clashed with giants such as Tencent and Alibaba in the ongoing battle for user traffic. If ByteDance uses this "hand of God" to leverage its "family bucket" of super app scenarios in collaboration with smartphone manufacturers, creating a super entry point that spans multiple applications, this move inherently challenges the current traffic allocation system.

Clearly, this business model poses a direct threat to traditional internet giants. Take WeChat as an example: its mini-program ecosystem has already formed a closed loop system independent of the App Store, posing a potential threat to Apple’s control over payments and distribution. Meanwhile, the cross-app operational capabilities of Doubao Assistant could precisely "break through" these ecosystem walls.

Given the current mobile ecosystem landscape, the difficulties and resistance Doubao AI phone is encountering are unsurprising.

So, do these "ecosystem walls" treat every player equally?

Suppose Tencent—which holds China’s largest social ecosystem—decides to enter the AI phone arena directly. How would it break through, or even turn these barriers into its own moat? Let’s venture a bold "simulation".

Who would be Tencent’s "ZTE"?

It’s highly unlikely that Tencent would choose to deeply collaborate with top-tier manufacturers such as Xiaomi, Huawei, or OPPO/Vivo, whose ecosystems have become very strong, because these companies would not readily give up control of their systems.

After all, these giants are no longer just hardware companies—they have each built vast ecosystem empires: Huawei has the HarmonyOS ecosystem and HMS, Xiaomi offers the AIoT network and Mijia, while OPPO and Vivo have invested heavily in ColorOS and internet services. Asking them to surrender control of their systems would be tantamount to shaking the very foundation of their existence, handing over their "thrones" willingly.

They stand as equals with Tencent, or even as potential competitors, and would never willingly become mere contract manufacturers or hardware affiliates within the Tencent ecosystem.

The best option for Tencent might still be to partner with manufacturers like ZTE—companies with strong manufacturing capabilities but relatively weak influence in the ecosystem—or even emerging brands eager to break through by differentiation, essentially those “muscular but landless” players.

For example, Lenovo could be a promising candidate, much like how Seres (formerly Sokon Group) has become a key partner for Huawei’s automotive ecosystem.

These players possess mature R&D, manufacturing, and supply chain capabilities, ensuring hardware quality and reliable delivery. However, when it comes to software ecosystems and user mindshare, they have yet to establish their own “moats.” For them, collaboration with Tencent isn't simply a business negotiation—it's a strategic lifeline for survival. Rather than struggling in a red ocean, partnering with a tech giant could gain them a differentiated ticket to the AI era.

Some emerging brands hoping to leapfrog the competition are also likely Tencent targets. With no historical burden and no ecosystem constraints, their greatest aspirations are to “be seen” and “be given uniqueness.” To them, Tencent’s AI ecosystem and massive traffic are incredibly tempting prospects.

Tencent’s cooperation model may also shift towards a “deeply integrated” approach.

It won't be satisfied with merely pre-installing apps—instead, it will seek capital-level involvement, using strategic investments or even acquisitions to fully align both parties’ interests. This would allow Tencent to influence every aspect, from hardware design and chip selection to OS customization and AI assistant integration, ensuring its vision permeates the entire process.

Who will play the role of AI assistant? Will “Hunyuan” take the spotlight, or will “WeChat AI” emerge as the new star?

Although Tencent has its powerful in-house Hunyuan large model, simply packaging it as a standalone "Hunyuan AI Assistant" on a phone’s home screen—like Siri—would not be the optimal solution and would squander one of Tencent’s most valuable assets. A far more compelling and disruptive move would be to launch a “WeChat AI Assistant” or similar product, deeply integrating WeChat’s capabilities.

We can imagine an experience in which “AI is WeChat, and WeChat is AI”—blurring the lines between the two in a seamless and intuitive way.

When you say, "Remind me to pick up my wife's favorite cake after work," the AI assistant will not only set a reminder, but also take into account your location and spending habits. It might even ask your wife about her preferences via WeChat, and instantly generate a card with navigation and recommended stores. When you say, "Send a 200-yuan red envelope to my colleague and wish him a happy birthday," the AI assistant will execute the task directly within WeChat. It could even craft a more personalized birthday message based on your chat history with that colleague.

This kind of proactive intelligence—powered by your social network, payment habits, and service history—is something no standalone AI assistant could hope to match.

It bypasses the long, difficult process of rebuilding user trust and habits—after all, WeChat has already become a "digital organ" for Chinese users, who rely on it daily for social interactions, work, payments, and numerous services. Injecting AI capabilities into this deeply trusted ecosystem means users face almost zero barriers to adoption. It doesn't feel like a "new feature you need to learn to use"; instead, it's the experienced "old hand" that seamlessly fits into all kinds of situations.

Competing companies might be able to replicate the algorithms, but they cannot copy the vast social network relationships, payment data, and extensive mini-program ecosystem that WeChat has built up over more than a decade. This so-called "WeChat AI" assistant will effectively become an "operating system within the operating system"—all services and information flows will eventually converge on this AI hub with WeChat at its core.

The Hunyuan foundation model is likely to serve as the brain and engine behind the scenes, providing the underlying technical support. In this vision, Tencent will wrap the core capabilities of Hunyuan in the familiar face of "WeChat AI", ultimately creating an AI super-app that seamlessly integrates social, services, and intelligence—an offering that can't be easily shaken. This approach will shift the competition over AI-powered smartphones from purely a technological race back to Tencent's strongest dimension: its ecosystem.

How will system-level permission calls be handled?

Confronted with the system-level permission challenges that have stymied competitors like Doubao, Tencent may opt for the elegantly simple approach of mini-programs.

The brilliance of this strategy is that it sidesteps the kind of confrontational thinking associated with "breaking through" at the app's core level. Instead, it builds a user-centric, highly compatible service coordination hub.

Doubao’s method of integration is more like that of a blunt intruder, trying to pry open every app’s front door with a universal key. Naturally, this triggers alarms—but ByteDance really has no other choice.

Tencent’s strategy, on the other hand, is a much more “gentle” approach.

When Tencent’s AI assistant needs to access third-party services, it typically doesn’t forcefully intervene or simulate user actions. Instead, it sends a command to WeChat, the “central system.” WeChat then functions as a dispatcher, seamlessly launching the relevant mini program within its ecosystem. Throughout the process, users are barely aware of any app switching or permission granting. This smooth experience reduces user anxiety over security and preserves a unified, fluid interaction.

For third-party apps, Tencent offers ample “respect.” When the AI assistant accesses a service, it clearly announces, “Currently using XX service for you,” which serves as an official endorsement and a source of traffic. They also gain real “benefits”: there’s no need to worry about core app data security being compromised, since all interactions occur within a controllable mini program sandbox. At the same time, they receive highly targeted and high-conversion traffic from AI-driven scenarios—something traditional advertising just can’t match.

This is a controlled cooperation. Apps relinquish some degree of control over interactions in exchange for a ticket to survive and thrive in the AI era.

How will those household-name apps respond? Will it be rejection or acceptance?

For many apps, WeChat is already much more than a simple social platform; it’s “infrastructure” vital for their survival. From user logins (WeChat authentication), viral growth (sharing to Moments), to closed-loop transactions (WeChat Pay), their lifeblood is deeply tied to the WeChat ecosystem. Offending Tencent would be tantamount to cutting themselves off from water and electricity supplies.

If Tencent makes even a slight adjustment to the weight it gives shared links, or introduces just a minor hiccup in the login interface's stability, it can instantly flatten an app’s user growth curve and send customer acquisition costs soaring. In this kind of "landlord and tenant" relationship, the tenant has almost no leverage to bargain with the landlord.

Although, given Tencent’s corporate culture, it’s unlikely they would act in such an unseemly fashion, as long as the possibility exists, there’s never a 100% guarantee of safety.

Tencent’s powerful business and ecosystem management abilities ensure that this will be a precisely orchestrated "carrot and stick" campaign.

Unlike Doubao, which charged in head-on, Tencent will, in all likelihood, launch an extensive "ecosystem partner briefing" well in advance. They’ll present highly enticing "carrots"—for example, promising apps with deep integration a "top-level recommendation" spot in its AI Assistant, providing access to richer data interfaces to help optimize their services, or even coordinating joint marketing with Video Accounts, Official Accounts, and other resources.

At the same time, that invisible "stick" will always loom overhead: if you refuse, your competitor could seize the opportunity, becoming the AI ecosystem’s "exclusive designated partner" and enjoying all the rewards this new era brings. In this atmosphere of collective anxiety—"join or be left behind"—few companies are likely to withstand the pressure.

Meanwhile, Tencent may also offer a “fair” choice: do you allow the AI Assistant to directly call your app, or would you prefer it to prioritize your Mini Program?

Choosing the former means granting Tencent deeper access to your data and control, weakening your own independence; choosing the latter means fully embracing the WeChat "Mini Program ecosystem," and effectively internalizing your service as a module within Tencent—users’ data and interactions would all be subsumed within the WeChat system.

No matter what choices are made, these apps have all transitioned from independent kingdoms of the past to becoming “vassal states” within the AI ecosystem. In the end, what appears to be a mutually beneficial cooperation is, in essence, a further strengthening of the ecosystem and yet another surrender of autonomy by numerous apps in the AI era.

For proprietary services within the WeChat ecosystem—such as WeChat Pay, Channels, Tencent Video, QQ Music, and the newly acquired Ximalaya—the intermediary layer of mini programs may even be completely stripped away.

AI assistants can achieve deeper, native-level integration, enabling direct access to core payment interfaces, media playback capabilities, and data pipelines. This means that when users tap into services, response times are faster, feature integration is more seamless, and the experience far surpasses that of any external mini program.

This differentiated design, in effect, establishes a “moat” for user experience, subtly steering users toward ’s own services and further consolidating and expanding its ecosystem empire.

If an AI phone becomes reality, its market response will most likely be a stark contrast to that of the Doubao AI phone. On the user side, powered by the trust and habit built by WeChat, a phone capable of delivering “AI is WeChat, WeChat is AI”—provided the experience is smooth and the functionalities are practical—will see far greater acceptance than any outsider. On the industry side, this will prompt a deep reconsideration of strategies. When the path of “forcibly breaking through” system barriers proves untenable, will seeking alignment and cooperation with existing ecosystem giants become the mainstream approach for launching AI phones?

This is merely a thought experiment—posing the question “what if”—where our perspective shifts the competitive focal point of AI smartphones from simply “algorithmic power” and “computing power” to a new battlefield of ecosystem rights. In the future, will AI smartphones evolve into OS-level AI-native devices, or will they trend toward an AI ecosystem housed within a super app? On the surface, these two paths seem to lead to the same destination, but in reality, they will define the power structure of the future mobile internet.

Will the “walled gardens” become even more impenetrable, or will the ongoing wave of technological advancement give rise to new models of collaboration and symbiosis?

This battle for ecosystem rights—one that affects every player at the table—has only just begun.