Reporter Says Muse Read His Messages; Meta Says Two Settings Must Be On

A claim that Meta’s personal AI, Muse, read private messages without permission has drawn a rebuttal from the company. On September 30 local time, TechCrunch reported both the complaint raised by reporter Jason Aten and Meta’s response. The crux of the dispute: when Muse made use of the contents of his Mac’s Messages app, had the user granted access permission or not?

Were message permissions turned on?

Aten says he never granted access to his messages, yet Muse sent him a notification based on the contents of his conversations. He also claims that macOS “Full Disk Access” was switched off. He wrote that he saw a record on his device showing that message contents had been synced.

Meta’s explanation is that reading messages requires specific conditions to be met first. Andy Stone, Meta’s vice president of communications, said that for Muse on Mac to access messages, the user has to turn on two settings. One is macOS “Full Disk Access,” and the other is the messages connection feature inside Muse. The reporter’s account and the company’s account are in direct conflict over whether permission was ever granted.

Muse’s own explanation of how it accessed the conversation also became part of the controversy. Muse told Aten that it had learned the contents by seeing a notification banner. Meta executive David Singleton countered that this answer was also wrong. That leaves two issues on the table at once: how Muse actually obtained the conversation content, and whether it described that route accurately.

What does Muse actually do?

Muse, which Meta unveiled in September 2026, is a personal AI that carries out tasks directly inside connected apps. It was introduced not merely as something that answers questions but as an AI that gets work done on its own (an agent), sending emails or booking travel. Meta said users can decide for themselves which apps to connect and how much access to grant. This controversy hinges precisely on how a user’s choices are applied during that app connection process.

Back in July, Meta had already announced planning capabilities, email and calendar connections, and task execution for its existing Meta AI. Muse, which came later, was presented as a personal agent equipped with a cloud virtual computer and browser provisioned separately for each user. The pitch centered on both the idea of working inside apps on the user’s behalf and the architecture of giving every user their own workspace.

Is it available in Korea?

Korea is not among the regions Meta has listed. In its launch announcement, Meta said Muse would roll out in stages in the United States across iOS, Android, and the web. In a follow-up announcement covering features for small businesses, it listed availability as the United States and Canada.

In that same small business announcement, Meta said Muse can be used free of charge for most purposes. It added that a subscription plan would be offered for users who want to use it more heavily.

The heart of this dispute is whether an AI that acts on a user’s behalf stays within permitted boundaries when it handles personal information. With services like Muse that work by connecting to apps, choosing which apps to connect is inseparable from whether those permissions are actually enforced. What remains to be verified: evidence showing the permission settings at the time and the route by which the messages were accessed, along with findings from an investigation that can reconcile the conflicting accounts.