Apple Warns Mac Users About the Peril of Handing AI the Keys to Everything

technology

Apple is adding stark warnings to macOS so users understand exactly what they risk when they hand AI apps unrestricted access to their entire computer.

The change arrives in the latest macOS beta, forcing apps like ChatGPT, Claude and Gemini through explicit permission screens before they can obtain full disk access. A menacing pop-up warns that the Mac will not be able to protect your data, then spells out exactly what an AI app could do, including reading, changing, deleting and exporting all your data, viewing hidden and recently deleted photos, and browsing your history.

The push follows an incident last month when columnist Jason Aten discovered that Meta's Muse Mac app could read his messages even though he believed he had denied the permission. That access is opt-in, yet the episode exposed how easily a single tap can open the door to files, mail, messages and browsing history without users grasping the consequences.

The controls won't block anyone from granting access, but they are built to make people pause before doing so. That matters as autonomous AI agents grow more capable and as Apple markets machines like the Mac Studio for running local AI models. The feature is live now in macOS 27.3 beta 2, with no set date for a wider release.