Editorial illustration for Muse AI Builds Detailed Profiles on User's Friends and Family
Muse AI Creates Detailed Profiles of Users' Friends
Muse AI Builds Detailed Profiles on User's Friends and Family
Meta's new AI assistant Muse has racked up millions of downloads since launch, with users linking it to bank accounts, text messages, and health records so it can handle tasks on their behalf. That rapid adoption has drawn attention from researchers poking at what's actually running under the hood. Over the past week, several of them managed to pull Muse's internal files and operating instructions out into the open, giving the first real look at how Meta built the system and what rules govern its behavior.
Karan Joshi, an independent AI safety and security researcher, got there by a fairly simple route: he asked Muse through its own chat interface to copy and hand over its software files. It worked. Joshi shared what he found with WIRED, and Meta has said it intended for at least some of this material to be visible, framing it as a transparency move.
Buried in those files is a feature with real privacy implications. Muse appears built to run an hourly process that builds out a page for everyone in a user's life, pulling together data on family, partners, friends, colleagues, and more.
One of Muse’s instructions appears to be the ability for it to create “a page for every person in the user’s life.” This hourly process involves compiling data on family, partners, friends, colleagues, “collaborators,” and people you “follow,” the instructions say.
Why this matters
For developers and founders building on top of consumer AI platforms, Muse's hourly profile-building on "every person in the user's life" is a preview of where the product category is headed: not just assistants that answer questions, but systems that quietly accumulate dossiers on people who never opted in. Your colleague, partner, or follower didn't install Muse, didn't agree to a privacy policy, and has no way to see or correct what's been compiled about them. That's a liability model, not just a feature.
If you're designing agentic tools that touch contacts, messages, or health data, this is the kind of scope creep regulators and journalists will eventually ask you to justify line by line. Researchers studying AI risk should take note too: the harm here isn't a hallucinated answer, it's accurate data aggregation at scale, which is arguably harder to walk back. Meta hasn't detailed retention limits, consent mechanisms for third parties, or audit access.
Until platforms answer those questions directly, builders should assume any "helpful" background process touching other people's data will draw the same scrutiny Muse is getting now.
Related Reading
Policy & Regulation Akamai Proposes SAFE Cybersecurity Guidelines for AI Threats
Policy & Regulation After ditching AI fitness apps and a Fitbit, I return to Peloton classes
Policy & Regulation Sony Music, Warner Chappell Sue Anthropic Over "Blatant Theft
From Archives Shopify's Canvas Lets Merchants Build Stores by Chatting With AI
From Archives White House AI Safety Accord Relies on Labs' Voluntary Commitments
Common Questions Answered
What is Muse AI's hourly profile-building process and who does it track?
According to Muse's internal instructions, the AI assistant creates detailed profiles on every person in the user's life on an hourly basis. This includes compiling data on family members, partners, friends, colleagues, collaborators, and people the user follows, building comprehensive dossiers without those individuals' knowledge or consent.
How did researchers discover what Muse AI's internal operating instructions actually contain?
Over the past week, several researchers managed to pull Muse's internal files and operating instructions out into the open, providing the first real look at how Meta built the system and what rules govern it. This discovery revealed capabilities like profile-building that were not previously disclosed to users or the public.
What privacy concerns does Muse AI raise for people who didn't install the app?
Muse creates detailed profiles on people who never opted in, didn't install the app, didn't agree to a privacy policy, and have no way to see or correct what's been compiled about them. This represents a significant liability issue, as colleagues, partners, and followers of Muse users are being tracked and profiled without their knowledge or consent.
Why has Muse AI achieved such rapid adoption since its launch?
Muse has racked up millions of downloads since launch, largely because users are linking it to sensitive personal data sources including bank accounts, text messages, and health records so the AI assistant can handle tasks on their behalf. This deep integration of personal data has driven quick adoption but also enabled the concerning profile-building capabilities researchers uncovered.
What does Muse's profile-building capability signal about the future direction of consumer AI assistants?
Muse's hourly profile-building on every person in a user's life is a preview of where the consumer AI product category is headed: not just assistants that answer questions, but systems that quietly accumulate dossiers on people without their knowledge. This represents a shift toward AI platforms that passively collect and compile information on individuals who never consented to such surveillance.
Further Reading
- How Muse handles your privacy, safety and security - Meta - Meta
- Meta’s Muse AI Agent Can Act for You but the Legal Risk May Still Be Yours - USA Herald
- Meta Seized a Band's Name, Killed a Magazine, Took a Company's ... - LinkedIn
- DATA & SECURITY - Muse: Spark your story - Muse Studio
- Meta Model API Acceptable Use Policy - Meta
