Meta disputes claim that Muse read a user's private messages without permission

Meta is pushing back hard against claims that its Muse AI agent snooped through a user's private messages without permission, according to reports from multiple tech outlets. A journalist said Muse accessed his private chats even though he had disabled the Mac system setting that should have blocked it. Meta says that's technically impossible.

The company insists Muse cannot read Messages content unless a user explicitly grants access through macOS permissions. If the setting is off, Muse should be completely locked out. But the journalist's experience suggests either a bug in the permission system, a misunderstanding of what was actually accessed, or something more concerning about how these AI agents interact with operating system controls.

This isn't just a he-said, she-said tech dispute. It cuts to the heart of how much we can trust AI assistants that increasingly want deep access to our personal data. These agents promise to be more helpful by reading your emails, messages, and documents, but that value proposition collapses if users can't reliably control what gets accessed.

The timing is notable. Meta and other big tech companies are racing to ship AI agents that feel truly useful, which means asking for broader permissions than traditional apps. Apple, Google, and Microsoft are all pushing similar assistants that want to read across your digital life. The permission models weren't designed for AI that can understand and remember everything it sees.

What makes this particular case tricky is that Mac permissions can be confusing even for technical users. It's possible the journalist thought a permission was off when a related permission was actually enabled, or that Muse accessed metadata rather than message content. Without more technical detail, it's hard to know exactly what happened.

Meta has every incentive to get this right. Trust is fragile with AI tools, especially after years of privacy controversies around Facebook and Instagram. If users believe Muse is reading their private messages without permission, adoption will tank no matter how useful the agent might be.

What this means for you: Before connecting any AI assistant to sensitive data sources, audit exactly what permissions you've granted in your system settings. On Mac, check System Settings > Privacy & Security > Automation and App Management to see what each AI tool can actually access. And here's a useful prompt when trying a new AI agent: "List all the data sources and permissions you currently have access to on my device, and explain what you can and cannot see." Most assistants will give you a straight answer about their current access level.