Meta’s Muse Builds Shadow People Pages

A new report says Meta’s Muse builds “people pages” about your friends and family—even if they never used the service—raising fresh consent and privacy questions supported by Meta’s own policies.

Story Snapshot

A report describes Muse creating pages on people in a user’s life, including nonusers.
Meta’s policies confirm data lives in a user’s virtual machine and can include info about others.
Meta says nonuser information may come from public sources or what users share, and is not ad-linked.
Key claims like hourly refreshes lack independent technical verification in public records.

What the report claims Muse does with people in your life

Sakutto.ai reports that Muse creates “a page for every person in the user’s life,” such as family, friends, colleagues, and even followed accounts. The article says these pages can exist for people who never used Muse. It describes sections like Facts, History, Relationship, and more, and claims the pages refresh hourly. The underlying internal instructions are not provided for public review, so the exact wording and authenticity cannot be independently checked.

Meta launched Muse as a personal agent that can access user data and connected tasks. Meta’s announcement says people choose how much access Muse gets, including links to email or files. The Associated Press reported that Muse runs in a dedicated virtual machine that holds the agent and the user’s data. Those details help explain how a user’s agent could store information about others inside that private workspace.

What Meta’s policies actually say about nonusers

Meta’s Muse Privacy Policy says Muse processes things you share and things you create with the agent, inside your own virtual machine. The policy says Meta may combine information across profiles and devices in the same Accounts Center. Meta’s help page for people who do not use Muse says information about you may come from what a Muse user shares or from public sources. It also says this nonuser information is not linked to a Meta account.

Meta also states that each user’s virtual machine is isolated. That means another user’s agent should not access it. Meta says Muse does not share conversations or virtual machine data with its ad systems. These claims, if accurate in practice, would limit how far nonuser information travels beyond the user’s own workspace. They would not, however, resolve consent questions for the people described inside that workspace.

Where the evidence is strong—and where it is not

The strongest point is that Meta’s own materials confirm Muse can process and store user-supplied information about other people, including nonusers, in a user’s virtual machine. The weaker point is the hourly-refresh claim, which relies on a secondary report without released logs, code, or files for independent testing. The record also does not show company-wide dossiers about nonusers or direct collection from nonusers’ devices. That distinction matters for both policy and law.

Past research on “shadow profiles” shows how data from some people can reveal things about others who never opted in. That pattern raises concerns across the political spectrum. Conservatives see another case of tech giants building files without clear consent. Liberals see power tilting toward large firms over individual rights. Both sides see a system where people’s lives are recorded and inferred first, and asked about later, with limited tools to opt out or challenge use.

Why this matters for families, workers, and citizens

Everyday choices now feed software agents, not just apps. When one person tells an agent about a coach, neighbor, or coworker, parts of that person’s life may be stored and reused. Meta says users can inspect and download Muse files, but the help pages do not offer a clear path for a nonuser to see, fix, or remove what someone else entered about them. That gap fuels fears that tech rules protect platforms first and people second.

Law and policy are behind the curve. Courts and regulators are still sorting out how privacy applies when information flows through agents, not just websites. Legal analysts warn that agent data can be broad, sensitive, and attractive to advertisers, litigants, and law enforcement. That mix raises stakes for consent, retention, and training use. Meta says training is configurable, and ads do not use Muse data, but many readers will want proof and independent checks.

What to watch next

Watch for independent technical tests that try to reproduce “people pages,” measure update timing, and track what leaves the virtual machine. Look for clear processes that let nonusers request access, correction, or deletion of entries about them. Expect questions from lawmakers in both parties, since consent and control now affect everyone, not just users. Clarity on data training and cross-product combining inside Accounts Center will be key for trust.