Users of ByteDance's agentic AI phone Nubia M153 report that its Doubao assistant is blocked or limited by Tencent's WeChat, Alibaba's Taobao, and other apps
Cissy Zhou / Nikkei Asia :
Context & Ripple Effects
The reports arrive days after ByteDance put its Doubao voice assistant on ZTE's Nubia prototype, making the handset an early test of whether an AI assistant can act across China’s major app ecosystems. That initial Nubia M153 rollout shifted Doubao from a standalone model product toward an assistant-distribution experiment.
The friction also extends a longer ByteDance–Tencent access dispute: ByteDance previously challenged Tencent over WeChat and QQ access restrictions tied to Douyin content. The new point of contention is not content sharing but whether an assistant can operate inside rival apps.
First-order effects
- Nubia M153 users face a less capable Doubao experience in WeChat, Taobao and other affected apps, limiting the phone’s cross-app automation proposition.
- ByteDance’s hardware-assisted distribution test becomes dependent on permissions controlled by Tencent, Alibaba and other app operators rather than on Doubao’s model capabilities alone.
Second-order effects
- The limits make app-level integration and permission policies a competitive control point for AI assistants, potentially encouraging ByteDance to prioritize workflows it can complete outside rival platforms.
- Tencent and Alibaba gain an incentive to define more tightly what third-party agents may do in their services, while users may have to choose between assistant convenience and full access to incumbent apps.
Third-order effects
- If this pattern persists, the assistant-distribution layer will be shaped less by the handset or underlying model and more by incumbent platforms’ control of high-frequency workflows.
- The episode points to a more fragmented agentic-work surface in China, where cross-app assistants may require negotiated access rather than assuming the interoperability available to ordinary users.
The trend: Agentic AI is moving from model competition into a contest over who controls the app permissions and workflows that make assistants useful.