AI
Meta's Muse AI Builds Comprehensive Dossiers on Your Social Connections
Researchers uncovered that Meta's viral personal assistant compiles detailed profiles of every person in your life—raising fresh concerns about relationship data collection.
WIRED — AI
Meta's newly launched Muse personal assistant has drawn millions of users willing to grant it access to their banking, messaging, and health information. Now, internal documents extracted by researchers reveal the system's infrastructure for systematizing information about your social world—a capability that goes beyond typical AI personalization and warrants closer scrutiny.
How Muse Catalogs Your Relationships#
Security researchers recently obtained Muse's internal system prompts by asking the agent to share its own code through its standard chat interface. These files outline a routine process, running hourly, that creates individual profiles for everyone in a user's orbit: family members, romantic partners, friends, work colleagues, and people followed on social platforms.
The profiles function as a memory system. Muse gathers available data—both what users explicitly share and information accessible through connected services—and assembles it into sections like Facts, History, The Relationship, Common Ground, Unresolved Threads, and Strengthening. Over time, these pages accumulate details: residential location, occupation, recurring life events, birthdays and anniversaries, shared goals, and even personality notes about how someone "acts with each other."
The Relationship Intelligence Angle#
According to the extracted instructions, Muse aims to generate actionable relationship guidance. A user interested in strengthening a bond might receive recommendations—a friend's birthday to remember, a past conversation worth revisiting, a gesture of support tailored to that person's needs. The system is designed to mimic how a close friend might remember your social landscape.
"What it seemed like—from all these prompts and skills data being fed into Muse—is that they want to understand your relationships with real people," says independent AI safety researcher Karan Joshi. "They're trying to know you like a friend, which is honestly pretty creepy."
Privacy Architecture and Transparency Claims#
Meta argues the approach is reasonable. Each Muse user operates within a dedicated virtual machine inaccessible to other users or agents, and users can delete memories or disconnect external services at will. The system also requests human approval before sending emails or authorizing purchases, and maintains an audit log of all activity.
A Meta spokesperson emphasized that relationship insight requires context: "For any agent to be useful and actually help you achieve your goals, it needs to have context about you and those you interact with. Muse gathers that based on public information and from what you've chosen to share."

