Apple announced changes to its macOS privacy settings on Friday aimed at restricting how third-party application developers can leverage full-disk access permissions. The move follows a significant public controversy stemming from an incident in which Meta’s Muse AI agent sent a notification to tech columnist Jason Aten referencing a private message thread, despite Aten not having knowingly granted the application access to his communications.
Meta’s Chief Technology Officer David Singleton responded to the initial uproar by stating that the Messages integration in Muse is opt-in and requires users to manually enable two separate permissions: a system-level full-disk access setting and a specific Messages connector configuration within the application. Singleton’s statement suggested that if Muse accessed messages, the user must have explicitly enabled both settings. However, macOS security expert Patrick Wardle countered this assertion, noting that full-disk access inherently permits applications to read any non-root files on a system, including messages and browsing history, regardless of additional settings.
Apple’s statement on the policy change emphasized concerns about developers utilizing full-disk access in ways that could expose users’ files, emails, messages, and browsing history without adequate awareness. The company highlighted particular risks associated with communication apps and stressed that as AI agents become more capable and autonomous, the security implications of granting such broad access will intensify. While Apple did not identify Meta or Muse by name, the timing of the announcement—following weeks of social media discussion about the incident—suggests a likely connection.
The announcement follows additional security revelations regarding Muse, including a reported vulnerability that could allow attackers to assume control of the assistant through ClickFix attacks, potentially accessing all resources available to the application. Additionally, Amazon had recently removed Muse from its platform, citing concerns about how such applications handle access permissions and respect service provider policies. Collectively, these developments raise questions about whether AI agents requiring such extensive system-level access are appropriate for general use without more robust safeguards.
Article Attribution | Read More at Article Source
Article summary produced by Claude AI