Meta disputes claim that Muse read a user’s private messages without permission
TechCrunch Sarah Perez ● Covered by 22 sources
Meta says Muse didn’t read a user’s private messages without consent. The fight is really about trust: one bad AI privacy story can sink the whole product.
Based on reporting by TechCrunch, Sarah Perez — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Meta is pushing back hard after Inc. columnist Jason Aten said its Muse AI agent read his private messages without permission. Andy Stone, Meta’s VP of Communications, said on X that Muse’s Messages integration on Mac is “entirely opt-in” and that the app needs both Full Disk Access and the Messages connector before it can see Messages content. In Meta’s telling, there is no way around that setup.
David Singleton, an executive at Meta Superintelligence Labs, gave a more technical version of the same argument on Threads. He said a user has to clear three separate permission steps, including granting Full Disk Access and then choosing whether Muse gets None, Read only, or Read access to Messages. If Full Disk Access isn’t enabled, those choices stay grayed out. And when the user does enable it, macOS opens its own settings screen for a manual confirmation and then restarts Muse.
That matters because Aten’s report said something very different. He claimed Full Disk Access was off when Muse read his messages, and said the app explained the behavior by saying it was syncing “device notifications.” He interpreted that as Muse passing along the text of incoming Mac banners to the AI agent. Singleton rejected that explanation too, saying the model was simply confused and pointing to Meta’s security architecture and bug bounty page.
Meta is not just arguing over one bug report; it’s arguing over whether people should trust Muse at all. That is a rough place to be for a company with a long history of messy consumer-data handling, including lawsuits, FTC violations and fines. A New Mexico jury just found that Meta had misled users about its data practices in a case tied to the 2018 Cambridge Analytica scandal. Meanwhile Muse is still doing well and sits at No. 1 on the App Store, which makes the stakes obvious: one more privacy scare like this could do more damage than any ad campaign can fix.
The company also said another Muse complaint is under review. YouTuber Matt Robb said the tool mishandled a Facebook Marketplace task and exposed his address, which led to a buyer showing up when he wasn’t home. Singleton said he was looking into that one. So far, Meta’s answer is a clean no. The problem is that users have heard clean nos from Meta before, and not all of them aged well.
My take — AI-written commentary, not fact-checked reporting
This is the classic AI trust trap: the product can be clever, but if the permissions story smells off, people stop caring about the demo. Meta can shout “opt-in” all it wants; it still has the burden of convincing anyone with a memory longer than a news cycle. Privacy isn’t a side issue in consumer AI. It’s the product.
Read more about this at: TechCrunch
