1 hour ago
5
Google’s Gemini Desktop app may soon get a lot more freedom on your computer. Its Tasks mode could reach into local folders and launch apps without asking permission at every step, according to TestingCatalog.
What Google is building into Tasks mode
The feature set falls under what Google calls “Computer Use.” It runs through the app’s Tasks mode.
The idea is to let the Gemini agent work directly on the machine it’s installed on. That means reading specific local folders, controlling desktop apps, and running multi-step jobs without stopping to ask for approval at each turn.
Signs of the capability first showed up in Gemini Desktop settings around August 18, 2026. Since then, developers appear to have been adding more detailed permission controls along with some safety features.
One of those features is an optional backup. Before a task starts, the app can copy selected folders to Google Drive automatically.
Users can also set app permissions individually, allowing or blocking Gemini’s access to particular programs.
A rollout taking shape
The Mac version of Gemini Desktop arrived earlier in 2026. The Windows version followed on September 11, 2026, opening the door to a much larger pool of potential users.
After the Windows launch, Tasks mode and the Computer Use features headed into further testing. By late September 2026, app control permission settings had reached trusted testers, which points to Google getting ready for a wider release.
As of October 2, 2026, reports indicate Google is still expanding those permission settings. Additional features reportedly in the works include a mini activity monitor and voice control through Gemini Live.
The competitive picture
Google isn’t building this in a vacuum. AI-assisted desktop automation has become a contested space, with OpenAI and Anthropic among the main players pursuing similar agent capabilities.
For Google, desktop apps on both Mac and Windows give Gemini a foothold outside the browser. That matters because so much of daily work still happens in local files and native programs.
What this means for users and the market
For everyday users, the near-term question is practical: how much access are you comfortable granting? The granular controls suggest Google expects people to answer that question in very different ways.
For businesses, an agent that can run multi-step workflows on employee machines could save real time, but IT and security teams will want clear visibility into what it touches and when. The mini activity monitor and per-app permissions read as early answers to those concerns.
Key things to track include how broad the default permissions turn out to be, whether the Drive backup ends up opt-in or opt-out, and how clearly the activity monitor reports what the agent is doing.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.
English (US) ·