Google is testing expanded capabilities for the Gemini Desktop application, specifically within its Tasks mode. The feature, referred to as Computer Use, aims to allow the AI agent to interact directly with local files and desktop applications without requiring constant user approval for every step.
The development of these controls began around August 18, 2026. Since then, Google has introduced safety measures, including an optional backup feature that copies selected folders to Google Drive before a task initiates. Users can also manage permissions for individual applications.
The Gemini Desktop app was released for Mac earlier in 2026, followed by a Windows version on September 11, 2026. By late September 2026, permission settings for app control were provided to trusted testers, signaling preparations for a broader release.
As of October 2, 2026, Google continues to refine these permission settings. Future updates may include a mini activity monitor to track agent actions and voice control integration via Gemini Live. These developments place Google in competition with OpenAI and Anthropic.
