手机智能体“无障碍”通行，你做好“直播生活”的准备了吗

Core Viewpoint - The article discusses the emergence and implications of AI agents in smartphones, highlighting their capabilities and the associated privacy risks due to the use of accessibility features [1][2][10]. Group 1: AI Agents in Smartphones - AI agents have evolved from simple chatbots to sophisticated systems capable of performing complex tasks, akin to autonomous driving technology [2]. - Major smartphone manufacturers are integrating AI agents into their devices, enhancing user experience by allowing tasks to be executed without switching between apps [2][13]. - Despite the advancements, current AI agents have not achieved full automation and exhibit vulnerabilities, particularly concerning privacy and security [10][12]. Group 2: Privacy and Security Concerns - The accessibility feature in smartphones grants high-level permissions, enabling AI agents to access sensitive information such as passwords and chat histories [3][4]. - There have been instances of misuse of the accessibility feature by malicious entities, raising concerns about data security [3][12]. - Testing revealed that most AI agents, except for Huawei, utilize the accessibility feature, with varying degrees of management and transparency [4][5][8]. Group 3: Recommendations for Improvement - The article advocates for enhanced management of accessibility permissions, including clear notifications when AI agents utilize these features and the ability for users to control their usage [14]. - Specific recommendations include listing AI agents in the accessibility feature settings, providing explicit consent prompts, and allowing users to disable the feature easily [14].