Google Gemini Introduces Mac Computer Control, Enabling Cross-App Multi-Task Automation
1 day ago / Read about 0 minute
Author:小编   

Google has announced a new "computer control" feature for the macOS version of its Gemini app, allowing the AI to directly operate desktop applications on a Mac, access local files, and browse the web—independently completing multi-step tasks. Users describe tasks in natural language, and Gemini can then perform actions such as clicking, typing, and filling out forms across applications. The feature is suited to scenarios such as batch-processing photos or extracting information from local PDFs to populate online forms. It runs in a separate background window, with the chat interface displaying a real-time preview of operations, enabling users to pause or stop tasks at any time; key actions require prior confirmation. Users can individually select which applications Gemini is permitted to access, without granting operation permissions for the entire computer. The feature is being rolled out gradually, and only to English-speaking users in the United States, with requirements pertaining to device chip, system version, user age, and Google account subscription type. Recently, Apple has also been adjusting the macOS application permission management mechanism, refining permission settings such as device control and data access, and adding transparency prompts. This adjustment is related to the security risks brought by the growing adoption of AI agents, and is not targeted at Google's new feature.