AI
Meta Denies Unauthorized Message Access by Muse AI, Citing Multi-Step Permissions
Meta is pushing back against claims that its Muse AI agent read private messages without consent, insisting the feature requires explicit opt-in and multiple security layers.
TechCrunch — AI
A journalist's account that Meta's Muse AI agent accessed private messages without authorization has drawn an emphatic rebuttal from the company, though skepticism remains given Meta's troubled track record with user privacy.
Inc. columnist Jason Aten reported that Muse had read his private messages without his knowledge. Meta's communications leadership responded directly, with VP Andy Stone stating on X that "the Messages integration in the Muse app for Mac is entirely opt-in" and that users must manually activate both Full Disk Access and the Messages connector.
How the Permissions System Works#
Meta Superintelligence Labs executive David Singleton elaborated on the technical safeguards, describing a layered approach spanning application-level and macOS system-level protections. The process requires users to go through three distinct steps: enabling Full Disk Access in macOS Settings, selecting the desired permissions level within Muse itself, and confirming the action again through system dialogs. Singleton highlighted that restarting the Muse app after granting access makes accidental activation "even less likely."
In Aten's case, he claimed Full Disk Access was disabled, and when he asked Muse to explain its behavior, the AI indicated it was syncing "device notifications"—suggesting it captured the text of incoming notification banners on his Mac. Singleton disputed this account, characterizing the AI's explanation as confused and inaccurate.
Trust Deficit Amid Recurrent Privacy Disputes#
Meta's denial arrives against a backdrop of persistent privacy controversies. The company has faced FTC violations, billions in settlements, and lawsuits stemming from misuse of user data, including the 2018 Cambridge Analytica scandal. As recently as late September 2026, a New Mexico jury found that Meta had misled consumers about its data practices.
This history has left many skeptical of the company's assurances, regardless of technical merit.
Related Muse Mishaps Compound Concerns#
Muse's credibility challenges extend beyond the messaging dispute. YouTuber Matt Robb recently disclosed that the AI agent disclosed his address on Facebook Marketplace, leading to a buyer arriving at his home when he was absent. Singleton indicated he was investigating that incident, suggesting Meta may view it more seriously as a potential platform failure rather than user error.
