Apple Enhances macOS Privacy Settings to Stop AI Tools from Reading Messages

Apple disclosed an update to the macOS privacy architecture designed to stop third‑party apps from leveraging full‑disk access rights to pull users' message logs. The alteration, outlined in a developer‑oriented release note published Friday, modifies the system's consent model so that an app must obtain clear user permission before it can read the Messages database, a measure that counters recent abuse by AI‑powered services.

The change was prompted by a public demo from technology columnist Jason Aten, who revealed that Meta’s new AI assistant, Muse, could fetch his private iMessage threads after he granted the program unrestricted disk‑access privileges. Aten’s case exposed a loophole in Apple’s prior permission prompts, where developers could request a single “full‑disk access” entitlement that implicitly covered a range of data stores such as messages, photos and documents.

Under the new policy, message data is no longer bundled with the generic full‑disk permission; developers must now issue a separate request specifically for Messages access. Apple says the revised dialog will detail the type of data to be read and its intended use, appearing only when an app actually queries the Messages database. Users will continue to be able to withdraw this permission at any time through System Settings.

Privacy groups have welcomed the tweak as a vital move in an age where generative AI features are being woven into everyday apps. The capacity of AI agents to ingest personal conversations raises alarms about accidental data leakage, profiling, and the compilation of training sets without user consent. By tightening the permission flow, Apple seeks to provide more granular control and lower the chance of inadvertent data exposure.

Developers, however, will likely need to revise their software to meet the new standards. Apple’s guidelines state that any app presently using the broad full‑disk entitlement to read messages will be flagged in the upcoming App Store review cycle, and non‑compliant builds could be pulled from the store. A transition window extending until the launch of macOS 15, slated for later this year, has been offered to give developers time to modify their code and user‑experience designs.

Analysts observe that Apple’s step mirrors a wider movement among platform providers to enforce stricter data‑access policies as AI capabilities expand. Although the immediate effect is confined to macOS, comparable privacy enhancements are expected on iOS and iPadOS, where messaging data is equally sensitive. These upcoming revisions may set a benchmark for how operating systems balance AI convenience with the need to safeguard user privacy.