Apple Updates macOS Permissions to Stop AI Assistants Misusing Full-Disk Access

Apple has announced modifications to its macOS privacy controls to prevent third-party applications from misusing full-disk access permissions to read personal message histories. The move specifically targets emerging security and privacy loopholes exploited by autonomous artificial intelligence agents, which critics argue pose unprecedented risks when granted expansive system-level privileges on personal computers.

The policy shift follows a high-profile controversy sparked by tech columnist Jason Aten, who revealed that Meta's new general-purpose AI agent, Muse, sent an unsolicited notification referencing a private discussion on Apple Messages. While Meta CTO David Singleton insisted that the integration is strictly opt-in and requires both macOS Full Disk Access and an internal connector toggle, security researchers pushed back. Prominent macOS security expert Patrick Wardle noted, "From a technical point of view, with FDA (full-disk access), any (non-root file), is readable, browsing history, browser cookies, chats, etc etc etc."

The public backlash highlighted growing anxiety over autonomous software assistants equipped with deep operating-system capabilities. Social media users and privacy advocates likened unchecked AI agents to dangerous power tools that can inadvertently expose confidential enterprise and personal conversations if left uncontained, prompting questions over whether ordinary consumers understand the ramifications of system-level permissions.

By restricting how applications handle sensitive chat databases even when granted broad disk privileges, Apple aims to restore granular sandboxing to macOS. The update signals a broader industry effort to establish defensive barriers against background AI scraping, as tech giants face mounting regulatory scrutiny over how emerging AI software handles private consumer data.