Apple Tightens macOS Disk Access as AI Agents Expand
Apple will add stricter controls to macOS Full Disk Access, responding to privacy risks created by increasingly autonomous desktop AI agents.
What happened
Apple said it will introduce additional controls around macOS Full Disk Access, a permission that can expose files, mail, messages and browsing history. The company said users who want to grant this unusually broad access will need to take a much more explicit action.
The announcement follows growing scrutiny of desktop AI agents. Recent reports questioned whether Meta’s Muse could access private messages on a Mac, while earlier security research highlighted risks in applications that can read or manipulate sensitive local data. Apple did not announce a new agent product; instead, it changed the operating-system boundary around what applications may do once users approve elevated permissions.
Why it matters
Full Disk Access was originally associated with legitimate utilities such as backup software. Agentic applications create a different risk profile because they can interpret information, make decisions and take actions across multiple applications. A permission granted once may therefore become a continuing channel for data collection or unintended operations.
Apple’s move is significant because it shifts responsibility away from users having to understand an opaque system setting. It also signals that operating systems may begin treating autonomous software differently from conventional applications, even when both request similar technical privileges.
The unresolved question is implementation. Apple has not yet detailed the new prompts, enforcement mechanisms or whether developers will need to redesign existing workflows. The change could improve trust in desktop agents, but it may also make powerful automation harder to deploy. For the broader market, the message is clear: access permissions designed for passive software are becoming inadequate for systems that act on a user’s behalf.
Uncle Cat take: Apple’s most important contribution here is not another privacy warning; it is acknowledging that “user consent” becomes fragile when an agent can continuously interpret everything behind one checkbox.