What impact did app restrictions have on early AI agent phone prototypes?

Question

Accepted Answer

App restrictions effectively disabled early AI agent phone prototypes.

After the “Doubao Phone” prototype was released and sold in a limited run in December, it initially handled simple voice commands across apps, including ordering food and comparing shopping prices. But days later, major platform owners such as Alibaba, Tencent and JD.com restricted the phone’s built-in assistant access to their platforms. This move effectively disabled the phone’s AI agent, and ByteDance then turned off the powerful tool in certain circumstances, including when payments were involved.
Restrictions by major app platforms cut off assistant access
The prototype relied on its built-in assistant to carry out tasks across multiple apps. When Alibaba, Tencent and JD.com restricted access to their platforms shortly after release, the assistant could no longer operate normally within those ecosystems. In practical terms, the phone’s AI agent functionality was stopped rather than simply limited.
The agent was turned off in specific scenarios involving payments
ByteDance responded to the blocked access by turning off the tool in certain circumstances. One explicitly mentioned case was when payments were involved, indicating that even where voice commands could be understood, the agent’s capabilities were curtailed when transactions depended on app platform access. This meant the prototype’s “agent” behaviour did not remain consistently active across all task types.
Early prototypes showed cross-app tasks before access was blocked
Before the restrictions took effect, the prototype could follow simple voice commands to execute tasks across apps. The examples given were ordering food and comparing shopping prices, which demonstrate that the agent could connect to app functions during the initial phase. The subsequent platform restrictions reversed that working pattern soon after release.
The broader issue was control over user interaction within app ecosystems
Industry commentary in the coverage pointed to a control problem: platforms want to keep direct contact with users, otherwise “they lose control to another party”. That framing explains why gaining broad access to apps owned by other companies was a sticking point for AI agent devices. As a result, early prototypes faced immediate operational barriers when app owners limited access.

Reading history

Hide →

This history is cleared when the page is closed.

What impact did app restrictions have on early AI agent phone prototypes?

Loading

Was this article helpful?

Helpful

Not helpful

Report an issue

2

0

0

Share with friends