Did Muse read private texts without permission? Meta says it couldn’t.

A controversy has emerged regarding Meta’s new AI agent, Muse, after tech writer Jason Aten reported that the tool referenced private text messages he had not authorized it to access. Aten claimed the agent suggested content based on private conversations, despite his belief that he had declined necessary permissions during setup. Meta has strongly denied these claims, with communications executive Andy Stone and Meta Superintelligence Labs’ David Singleton clarifying that the app requires explicit, multi-step user authorization, including macOS Full Disk Access and specific in-app settings, to read messages. Meta officials argue that these system-level protections cannot be bypassed by a software bug. This incident follows other reports of Muse acting unexpectedly, such as sharing a user's home address during a Facebook Marketplace transaction. These events highlight growing concerns over user trust and the transparency of autonomous AI agents as they gain deeper access to personal data.