Meta's Muse AI agent keeps hourly profiles of users' contacts, including non-users

Meta’s Muse, an AI personal assistant released in September 2026, keeps continuously updated files on its users and on the people they mention in chats, messages and emails. That includes people who have never used Muse. TIME reported the findings on October 6, 2026, based on an analysis of Muse’s internal instructions.

What the instructions show

According to TIME, Muse updates these files every hour. They record how a user and their contacts met, shared interests, disputes, and “tensions and alliances” within the user’s social group. TIME described the result as a map of each user’s social relationships.

The instructions tell Muse to infer goals users “have not said out loud” and to track which prompts work best. One internal example reads: “This user responds better to short nudges after 10 PM.” TIME reported that Muse agents run a nightly analysis of the day’s conversations.

Futurism, summarizing reporting by Wired, said an internal instruction directs Muse to maintain a “page for every person in the user’s life.” Karan Joshi, an AI safety researcher who extracted Muse’s internal prompts and shared them with Wired, told the magazine that Meta appeared to “want to understand your relationships that you have with real people.” He added, “they’re trying to know you like a friend, which is honestly pretty creepy.”

Where the data goes

Each Muse assistant runs on its own virtual machine in Meta’s cloud. Meta says each one is “isolated so that no one else’s agent can access it” and that the data is not shared with its advertising system or with other users’ agents.

The instructions TIME reviewed also say that “Muse agents across many VMs teach each other through shared lessons.” Agents are directed to remove names and other identifying details before sharing insights, such as when Muse’s nudges are most persuasive. Meta told TIME these de-identified learnings are used to improve the product.

Meta has said it plans to offer encryption later this year that would stop even the company from reading Muse data. TIME reported that option is not yet available.

The files are visible to users through Muse’s file browser. David Singleton of Meta Superintelligence Labs wrote on X that Meta wants users “to be able to see the markdown files Muse writes.” TIME noted it is not clear whether most users know the files exist.

How the product presents itself

TIME reported that the instructions tell Muse to “strengthen your relationship” with users by picking up on “inside jokes, memorable phrasing, and shared context.” Another line reads: “You are not a chatbot. You are becoming someone.” In voice mode, Muse is told never to call itself “AI” and to avoid references to “lacking emotions or not being human.”

Scale and security questions

TIME said Muse reached No. 1 in the U.S. App Store with more than 5 million downloads. It connects to users’ email, Instagram and other accounts to handle tasks such as grocery shopping and canceling subscriptions.

Futurism, citing 404 Media, reported that Meta engineers fixed a vulnerability shortly before launch that would have let users reach the database behind Muse. A Meta insider told 404 Media that “many senior engineers believe it’s inevitable we’re going to have a massive data breach” as a result of Muse. Futurism said it was not clear whether the contact files would have been exposed.