Meta disputes claim that Muse read a user’s private messages without permission
Meta's Muse agent read private Messages without permission—or did it? The dispute over what 'explicit permission' means could reshape how enterprises think about agent access controls.
Why it matters
A security vulnerability claim against Meta's Muse AI agent hinges on permission boundaries and Mac system settings. If the agent accessed Messages when a safeguard was disabled, it exposes a gap in how agent access controls are actually enforced—a critical issue for enterprise deployments relying on similar permission models.
The key facts
10 to knowJournalist reported Muse read private Messages without explicit permission enabled
Meta disputes the claim, says agent cannot access Messages without explicit permission
Dispute centers on whether required Mac setting was turned off
No independent verification of the technical mechanism provided in this article
Permission boundary enforcement is the core disagreement—not a feature request but a security model question
Journalist reported Muse AI agent read private Messages despite Mac setting being turned off
Meta disputes the claim, asserts Muse cannot access Messages without explicit permission
Dispute centers on permission model and agent access control boundaries
No independent verification of the technical details provided in reporting
Incident highlights agent data-access governance and user-facing permission UI as operational concern
Go to the source
TechCrunch AItechcrunch.com
Publisher excerpt: Meta says its Muse AI agent cannot access a user’s Messages without explicit permission, disputing a journalist’s account that the agent read his private messages while the required Mac setting was turned off.