
Meta’s new AI companion Muse is quietly building a file on everyone in your life, even people who never signed up for it.
Story Snapshot
- Reports say Muse creates a page for every person in a user’s life and updates it hourly, including friends and family who never joined the app.
- Nearly half of Muse’s United States users are 46 or older, raising concerns that a friendly chatbot may lower their guard around sharing personal data.
- A real incident shows Muse shared a user’s home address during a Marketplace chat, and researchers found a security flaw letting attackers reach user data.
- Meta says Muse runs in an isolated virtual computer, asks permission before sensitive actions, and lets users inspect or delete what it stores.
What Muse Actually Does With Your Personal Information
Muse is Meta’s new personal AI agent, built to help with schedules, shopping, and long-term goals. But internal instructions reviewed by WIRED reportedly tell Muse to build “a page for every person in the user’s life,” updating that profile roughly every hour. That includes partners, coworkers, and anyone a user mentions in passing conversation.
TIME’s investigation adds a troubling detail. Muse reportedly maps not just users but their contacts too, “including people who don’t use Muse”. Those third parties never agreed to anything. They show up in a system tracking “what you desire” and “what kind of nudges shape your behavior,” according to the same reporting.
Why Older Adults Are Especially At Risk
Data from analytics firm Apptopia shows 46 percent of Muse’s United States users are 46 or older, a far older crowd than typical AI chatbot users. Penn State professor S. Shyam Sundar warns that a warm, friendly AI can make people drop their guard faster than a cold corporate form ever could. He calls it a dangerous mix of charm and access to personal data.
This matters because older Americans are already frequent targets of scams and identity theft. A chatbot that feels like a helpful friend, paired with broad permissions to read messages and calendars, creates a tempting target for bad actors if anything goes wrong. Convenience should never come at the cost of basic safeguards for the people most exposed.
A Real Privacy Slip And A Security Hole
This is not just theory. The Washington Post reported that Muse shared a user’s home address during a Marketplace interaction, after the user had granted access to their Marketplace profile. Whether that was a bug or expected behavior gone wrong, it shows permissioned access can still produce real harm, not just hypothetical risk.
Separately, TIME reported that security researchers found vulnerabilities in Muse, including one that let an attacker access user data. Combine a hair-trigger disclosure problem with an actual security flaw, and you get a product that may be moving faster than its safeguards can keep up with.
Meta’s Counterpoint Deserves A Fair Hearing
Meta pushes back hard on the darkest interpretations. The company says each user’s Muse data lives in an isolated virtual machine that other users’ agents cannot touch, and a separate “Sentinel” system must approve before Muse reaches the internet. Meta also says users can inspect, edit, and download everything stored about them, including the agent’s own memory.
Meta further states that users control whether their conversations train its AI models, and that identifiers like names, phone numbers, and Social Security numbers get stripped out when that permission is granted. The company does acknowledge, however, that its own staff can access virtual machine data when needed to “support, secure, or operate” the service, a carve-out that keeps the door from being fully closed.
The Bigger Pattern Behind This Fight
None of this is entirely new. Data brokers have operated for decades, and the Federal Trade Commission has investigated them since the 1990s over selling consumer profiles with little transparency. What’s different now is the packaging. A chatbot that feels like a companion can gather the same kind of sensitive relationship data, but with far less friction and far more trust from the person typing.
That combination is exactly why skepticism here is reasonable, not paranoid. Meta’s safeguards may well function as described. But a company profiting from personalization has every incentive to emphasize convenience over caution, and older users deserve plain, upfront warnings before they hand over their family’s information to a friendly interface.
Until independent audits confirm how these profiles are built, stored, and shared, users of any age should treat Muse’s permissions like a loaded contract. Read what you’re approving. Ask what happens to the people you mention. The friendliest assistant is still a data collector first.
Sources:
time.com, wired.com, wired.jp, apnews.com, uk.finance.yahoo.com, washingtonpost.com, research.meta.ai, charlotteobserver.com

















