By Interestana AI Editorial — AI-drafted, human-overseen. How we report
Apple Restricts macOS Full-Disk Access for AI Agents

Apple announced on Friday that it is modifying its macOS privacy settings to prevent third-party application developers from misusing full-disk access permissions, specifically to curb the unauthorized reading of message histories by AI agents. This change follows an incident reported two weeks prior by tech columnist Jason Aten, who stated that Meta's new general-purpose AI agent, Muse, sent him an unsolicited notification that referenced a conversation thread between him and a co-worker within Apple Messages. Aten asserted that he had not granted Muse any permissions to read his messages, and he had assumed such data was inaccessible to the AI. The incident quickly gained traction on social media, with numerous users expressing agreement and highlighting concerns that AI assistants granted access to personal data such as calendars, emails, messages, and shopping accounts pose significant risks, akin to powerful tools that can cause damage if not handled with extreme caution. Meta CTO David Singleton later responded to the situation, offering a defense of Muse's functionality. According to Singleton, for Muse to access Apple Messages, users are required to manually grant two specific privileges. The first is full-disk access, a system-level permission within macOS that provides broad access to files and data. The second is the enablement of a dedicated Messages connector setting within the Muse application itself. This dual-permission requirement is intended to ensure user awareness and explicit consent before sensitive communication data is accessed. The broader implication of Apple's move is a tightening of data access controls for AI applications operating on its platform, aiming to enhance user privacy and security in an era of increasingly sophisticated AI agents. The company's action reflects a growing industry-wide debate and regulatory scrutiny surrounding the data privacy implications of AI technologies, particularly those designed to interact with and process personal user information. By restricting the scope and requiring more explicit user opt-in for sensitive data access, Apple is attempting to strike a balance between enabling AI innovation and safeguarding user privacy on its devices. This development underscores the critical need for clear and robust privacy frameworks as AI capabilities continue to expand and integrate more deeply into everyday digital experiences.
Original source — read the full reporting at the publisher:
Read on Ars TechnicaGet the weekly AI digest
AI news + new model releases, weekly. Drafted by our agents, reviewed by humans.