Apple to Modify macOS Privacy Settings Amid Concerns Over AI Access to Messages

Highlights

• Apple plans to update macOS privacy settings to curb unauthorized access to message histories by third-party applications.
• This decision follows allegations that Meta's AI agent, Muse, accessed a user's Apple Messages without explicit permission.
• Meta's CTO, David Singleton, previously stated that Muse requires both "Full Disk Access" and an enabled "Messages connector" to read messages.
• Apple indicated that some developers are exploiting "Full Disk Access" permissions, exposing sensitive user data without full user awareness.
• The company emphasized that as AI agents advance, the dangers linked to extensive data access will significantly increase.

Details

Apple is implementing changes to its macOS privacy configurations, specifically addressing how third-party applications utilize permissions to access message content. This move comes after a tech columnist, Jason Aten, reported receiving an unsolicited notification from Meta's AI agent, Muse, which referenced a private conversation he had on Apple Messages. Aten asserted he had not granted Muse permission to read his messages, sparking widespread discussion on social media about AI assistants and data privacy.

Meta's Chief Technology Officer, David Singleton, responded by explaining that for Muse to access Apple Messages, a user must manually enable two distinct settings: macOS system-level "Full Disk Access" (FDA) and a specific "Messages connector" within Muse. Singleton's statement implied that any such access would have been due to the user's explicit configuration. However, Apple's subsequent announcement, explaining its reason for modifying FDA permissions, countered this by stating that some developers are leveraging FDA in ways that could expose a user's entire system, including files, email, messages, and browsing history, often without the user's complete understanding. The company also highlighted that for communication applications, this could jeopardize the privacy of those communicating with the user.

While Apple did not specifically name Meta or Muse, the timing of its announcement, following significant public concern, suggests a connection to the incident. Apple's clarification appears to contradict Singleton's earlier assertion that Muse could not read message content without the connector being enabled. The company stressed its commitment to ensuring users fully comprehend the risks before granting such comprehensive access, enabling them to make informed choices regarding their personal data and privacy.

Why it matters

As AI agents become more sophisticated and operate with greater autonomy, the associated risks of broad data access are set to intensify. Apple's proactive measure aims to ensure users are fully aware of these potential dangers, empowering them to make informed decisions about their personal data and privacy. This development underscores the delicate balance between the evolving capabilities of AI applications and the imperative to protect user privacy, highlighting the growing responsibilities of technology companies in this domain. It also brings to the forefront the need for clear and transparent permission management in an era of increasingly integrated AI tools.