Apple to tighten Mac Full Disk Access controls over AI agent privacy fears

Stricter limits on deep system access

Apple is changing how macOS handles its most powerful data permissions to prevent autonomous AI agents from quietly scraping sensitive user files.

In a blog post published on Friday, 2 October 2026, the company announced it will introduce stricter controls for granting Full Disk Access, a setting that allows software to bypass standard privacy boundaries. Apple warned that some developers are using this permission in ways that expose everything on a user's system, including files, mail, messages, and browsing history, without their full knowledge.

The decision comes amid a growing debate over how much freedom conversational and agentic AI tools should have on personal computers. As these AI agents become more autonomous, Apple says the risk of them abusing deep system access is growing.

The Meta Muse controversy

The policy shift follows a public dispute over Meta's Muse AI agent. Jason Aten, a technology columnist for Inc. magazine, accused the Muse assistant of reading private messages on his Mac, even though he had not granted the application Full Disk Access.

Meta has strongly denied the accusation. Andy Stone, a spokesperson for Meta, stated on social media platform X that Muse cannot access Apple's Messages app without explicit permission. He explained that a user must manually enable both Full Disk Access and a specific Messages connector within the app for Muse to read those files, adding that the permission can be revoked at any time.

What changes for Mac users

While the details of that specific incident remain contested, Apple is moving to ensure that such access cannot be granted accidentally or without a clear understanding of the consequences.

Currently, Full Disk Access allows backup utilities, security software, and other deep-system tools to run effectively. But when granted to an AI agent designed to scan, summarise, and act on a user's behalf, it effectively hands the keys to a user's entire digital life to a third-party model. Apple noted that this does not just threaten the user's own data, but also compromises the privacy of anyone they communicate with, as AI tools can scan incoming messages and emails.

To address this, Apple plans to implement new prompts and stricter verification steps. The company has not yet detailed the exact design of these prompts or when they will roll out to Mac users, but it has committed to requiring very explicit user action before any application can gain full access to a system.