Apple to tighten Mac privacy controls against AI agents after Meta Muse complaints

Apple plans to implement stricter privacy controls for AI agents on its Mac computers following complaints that Meta Platforms' Muse may have accessed private messages. The new safeguards aim to ensure that users who grant an app full disk access do so with a clear, explicit action. Full disk access allows apps to access potentially sensitive information, emails, messages, and browsing history, facilitating the proper functioning of backup apps.

Apple acknowledged that AI agents have become increasingly capable and autonomous, and that the risks associated with this level of access will grow. The company plans to introduce additional controls to ensure users can only grant apps such extensive access with a clear, explicit action. This move comes as Meta faces criticism over the safety and privacy of its AI efforts, particularly after Muse was launched last month.

Users have raised concerns that Muse could access private information, particularly messages. Despite Meta's assurances that Muse is safe and secure, user complaints have emerged about the app's ability to access supposedly private user data. In July, Meta scrapped Muse Image following a backlash over privacy concerns related to the app's automatic linking to Instagram accounts without users' knowledge.

The issue of AI risks has gained traction in recent months, with Meta joining a list of high-profile companies, including Anthropic, Google, and OpenAI, in disclosing potential compromises of other websites by their AI models. Apple's Mac computer revenue surged 29% year-on-year in the second quarter, with shipments increasing 30%.