21世纪经济报道·2025-12-05 05:58Core Viewpoint - The article discusses the recent announcement by Doubao Mobile Assistant regarding the limitations imposed on its AI capabilities in three specific scenarios: score manipulation, financial applications, and gaming usage [1]. Group 1: Product Launch and Market Response - Doubao Mobile Assistant was launched on December 1, in collaboration with ZTE, using the Nubia M153 model as a customized technology demonstration device [3]. - The core selling point of the AI phone is its ability to automate daily app operations, such as ordering food, booking flights, and price comparison shopping [3]. - Within four days of its release, the AI phone sold out, with resale prices on second-hand platforms reaching as high as 12,900 yuan for a device originally priced at 3,499 yuan [3][4]. Group 2: Limitations and Restrictions - Users have reported using Doubao Mobile Assistant for activities like video watching on Pinduoduo to earn rewards and attempting to use AI for gaming in "Honor of Kings," but these functionalities have been restricted [9]. - In financial applications, Doubao Mobile Assistant has faced obstacles, with users unable to log in or complete payments on apps like Taobao and major banks due to security measures [10]. - The assistant's reliance on screen reading and simulated operations has led to security concerns, as many apps have not authorized this type of automation, resulting in restrictions on its capabilities [10][11]. Group 3: Compliance and Industry Impact - Compliance experts have raised concerns about the risks associated with the AI's screen reading capabilities, which bypass the sandbox mechanisms that ensure data isolation between apps [11]. - The ability of AI to replace user operations could potentially disrupt the genuine traffic and monetization models of apps, impacting user data integrity [11]. - Doubao is actively seeking deeper communication with application developers to establish clearer and more predictable rules, aiming to avoid blanket restrictions on the reasonable use of AI by users [11].