Gemini Spark Lets AI Access Your Mac’s Local Files, Not Just Cloud Docs

The real story in Google’s Gemini desktop update is not the AI itself but where it operates: directly on your local file system, not just inside a browser tab or cloud document. That shift in scope is what separates this from the Gemini features most users already know.

Google began rolling out a new version of its macOS Gemini app this week, introducing a feature called Spark. The update is currently limited to an early test release for Google AI Ultra members in the United States. A new Spark tab appears in the app’s sidebar, and that tab is the entry point for giving the AI actual instructions over local files.

The file management capabilities are specific enough to be worth spelling out:

• Sorting files from a downloads folder into organized subfolders
• Pulling figures from invoice files and building a budget spreadsheet in Google Workspace
• Following a schedule to keep that spreadsheet updated automatically

Privacy controls are built around folder-level permissions. Users designate which folders Spark can access, and those connections can be removed at any time. The assistant only sees what it has been explicitly pointed at, which matters when the feature is reaching into local storage rather than a sandboxed cloud environment.

Remote triggering is the more consequential part

Google says a future update will let users initiate these local tasks from a phone. The described scenario is asking the assistant, while away from a desk, to locate a file, extract specific data, and send it by email, with the Mac doing the actual work in the background. That would make the desktop app function more like a persistent local agent than a chat interface.

Whether that remote access model raises comfort levels or concern will depend heavily on how the folder permission system holds up in practice. The feature is early access for now, so most users will not see it for some time.