Apple announced modifications to its macOS privacy framework following growing concerns about how third-party applications, particularly AI agents, access sensitive user information. The move comes amid scrutiny over Meta’s Muse AI assistant, which reportedly accessed Apple Messages content without explicit user awareness, though Meta maintained the integration required deliberate user opt-in and specific permissions.
Meta’s chief technology officer argued that Muse could only read Messages if users independently granted both full-disk access at the system level and enabled a specific Messages connector setting within the application. However, security experts questioned this explanation, noting that full-disk access inherently permits reading of all non-root files, including messages, browsing history, and cookies. Apple’s statement suggested developers have been exploiting full-disk access permissions in ways that could expose users’ entire system contents without adequate awareness or consent.
Apple emphasized that as AI agents become increasingly autonomous and capable, the risks associated with granting such comprehensive system-level access will escalate significantly. The company stated its commitment to ensuring users understand these risks before authorizing such permissions. While Apple did not identify specific applications or companies, the timing of the announcement—occurring shortly after the Muse incident generated significant public discussion—suggested a connection.
The announcement followed additional security concerns about the Muse assistant, including a disclosed configuration vulnerability that potentially allowed unauthorized actors to commandeer the system. Amazon separately restricted Muse from its platform, citing concerns about whether such applications operate transparently and respect service provider policies.
These developments collectively raised questions about whether AI agents requiring extensive system access warrant the permission levels they currently request, with observers noting that users may require more substantial safeguards than existing permission mechanisms provide.
Article Attribution | Read More at Article Source
Article summary produced by Claude AI