Apple Tightens macOS Disk Access as AI Agents Increase System Risks

Apple is implementing new restrictions on the "Full Disk Access" setting within macOS, citing heightened security risks driven by the rise of autonomous AI agents. According to a blog post from Apple, the feature was originally intended to enable backup applications to function correctly, but some developers are now utilizing this access in ways that could expose a user's entire system without their full understanding.

Apple warned that granting this level of access allows an application to read browsing history, mail, messages, and files. The company noted that for communication apps, this access can compromise the privacy of the individuals with whom the user is communicating. Apple stated that as AI agents become more autonomous and capable, the associated risks will grow substantially, necessitating a shift toward requiring "very explicit user action" for those who wish to grant such extraordinary permissions.

Reporting from TechCrunch indicates the move follows several security concerns. Inc. columnist Jason Aten reported that Meta's Muse app on Mac was able to read the content of his private messages, despite Aten claiming he had not granted the AI agent permission. Meta has disputed this claim. Additionally, TechCrunch cited a Wired report stating that a vulnerability in the Mac application for ChatGPT could have enabled hackers to access sensitive data.

While the Muse app optionally allows users to enable Full Disk Access to provide the AI with greater reach into personal content and files, Apple's new controls are designed to ensure users make informed decisions regarding their data privacy. Apple did not respond to inquiries from TechCrunch regarding the specific details of the feature change.