New Google Gemini setting could give AI tool broad access to users' Macs
Google is testing a permission model that could let Gemini agents access your Mac files without asking every time — a convenience-vs-control tradeoff that enterprise IT teams need to watch.
Why it matters
A new macOS permission setting would allow Google Gemini to operate as an agent with persistent file and action access, bypassing per-action approval. This raises operational security and compliance questions for enterprises deploying agentic AI on employee devices.
The key facts
10 to knowNew macOS permission setting grants Gemini broad file and action access without per-request approval
Agent autonomy enabled through permission delegation rather than per-action consent
macOS-specific implementation; applies to Google's Gemini agent on Apple devices
No published safeguards, audit trail, or granular scope limits disclosed in reporting
Timing: October 2026; status appears pre-GA or limited rollout (setting described as 'new')
New permission grants broad file access and action execution without per-request approval
Applies to Gemini on macOS
Reduces friction for agentic workflows but increases risk surface if compromised
No pricing, rollout date, or scope (consumer vs. Workspace) disclosed in article
Permission model mirrors OS-level agent-access patterns emerging across platforms
Go to the source
ITProitpro.com
Publisher excerpt: A new permission would allow an AI agent to access files and carry out actions without specific permission every time