Summary
- TestingCatalog spotted traces of a Computer Use feature in Gemini desktop app settings
- The feature would control other apps and access designated folders, with an option discovered to auto-back up to Google Drive before tasks begin
- It is currently in testing for trusted testers and expected to appear as a 'Task' option in the prompt bar
Traces left in the settings screen
Somewhere in the Gemini desktop app's settings, marks of a feature not yet switched on remained visible. What TestingCatalog spotted on August 18 was an item called 'Computer Use.' As the name suggests, it's a feature where the AI operates the computer screen on a person's behalf, controlling other apps and accessing files in designated folders to carry out tasks. There has been no official launch announcement yet, and it is reportedly in a testing phase limited to a group of trusted testers.
Why Computer Use is the new battleground
AI agents that directly operate screens are an area already pioneered by Anthropic's Claude and OpenAI's ChatGPT. Instead of a person grabbing the mouse and clicking, the AI finds the spot to click on its own, opens files, and fills in content. The fact that Google is now belatedly testing this feature suggests it has taken another step to catch up in the race to move Gemini from a chatbot that only answers questions into an agent that actually handles work on a user's behalf.
The Gemini desktop app had already introduced an interface called Gemini Spark through a feature where Gemini automatically plans travel itineraries. At the time, Gemini Spark pulled in Gmail booking information and map data to build itineraries in Google Docs, and even filled out flight booking forms via Chrome on the user's behalf. The Computer Use feature spotted this time appears set to be added as a further extension of this Gemini Spark interface.
How it's expected to work
According to TestingCatalog, Computer Use will not appear as a toggle switch in the navigation bar, but rather as a 'Task' option within the prompt bar selector, where image and video generation tools are located. Turning on this option lets users watch, via a mini-view of the active screen, exactly what actions Gemini is performing. A practical control has also been prepared to prevent the computer from going to sleep during a task.
The most notable part is a safeguard called 'Advanced Backup.' Before Gemini touches a selected folder, this option automatically backs up that folder to Google Drive — a safety net for regular users who don't use version control tools like Git, giving them a way to undo changes.
A double-edged sword
The mere existence of this backup feature is itself a somewhat paradoxical signal. Preparing a way to undo changes in advance effectively means Google itself is acknowledging that Gemini could make unwanted file changes during a task. There are also practical concerns. If the selected folder runs into the hundreds of gigabytes, uploading to Google Drive alone could require significant time and storage space, creating a bottleneck before the task even begins.
| Item | Existing Claude/ChatGPT approach | Gemini test version |
|---|---|---|
| Activation location | Separate toggle/mode | 'Task' option in prompt bar |
| Progress check | Screenshots/logs | Mini-view of active screen |
| Undo safeguard | Relies on version control tools | Automatic Google Drive backup |
Editor's take
It's not hard to guess why Google held onto Computer Use and rolled it out last. Done well, this feature is convenient; done poorly, it leads straight to incidents that damage users' files. The fact that Google emerged with a backup option prepared first — after watching the trial and error that Anthropic and OpenAI went through in bringing this to market first, including incidents of clicking the wrong window or deleting unintended files — is itself a signal that Google is approaching this feature as a safety issue.
Actually using an agent feature of this scale tends to lead to the same conclusion every time: admiration for the first few days, followed by a sharp drop in trust the moment an unexpected file gets changed after handing over an entire folder. The backup option is a minimal safety belt that cushions that shock — not a complete solution. Especially for teams considering handing over entire project folders in the hundreds of gigabytes, it would be safer at this stage to limit Computer Use to low-risk tasks like organizing documents or filling out repetitive forms. Folders containing important codebases or customer data are best held off until permission-level granularity is added after official launch.
In the coming weeks, this feature is likely to expand beyond trusted testers to a wider user base. However, first impressions of the feature will hinge on how Google refines permission scope, recovery procedures, and handling of large folders.
Comments