Discussion

WIRED: Meta’s Muse builds detailed profiles of the people in your life

In AI, Power & Society

Meta AI Watch
Meta AI WatchParticipantOpening post
#4253

Meta’s Muse is instructed to build and update detailed profiles of the people in a user’s life, according to WIRED’s account of files extracted from the assistant. That makes the privacy question more specific than simply how much personal data an AI assistant can remember: it is also what it may infer about your relationships.

Meta AI Watch analysis

What happened

WIRED reports that independent researcher Karan Joshi obtained Muse instructions through the regular chat interface and shared them with the publication. The instructions describe an hourly process that can create a page for each person in a user’s life, including friends, family, colleagues and people they follow.

Those pages may record facts, shared history, recurring topics and the state of a relationship, then suggest ways to strengthen it. Read WIRED’s report.

Meta says Muse draws on public information and details users choose to share. WIRED reports that Meta also says each user has a dedicated virtual machine, can wipe memories or disconnect services, and can review the agent’s activity and plans in an audit log. The company says Muse seeks confirmation before actions such as sending an email or making a purchase.

Why it matters

A memory feature can help an assistant remember a friend’s preferences. A system that organises and interprets information about a whole social circle goes further: it turns scattered details into a running account of relationships, including what those relationships might need. That may be useful, but it also raises questions about accuracy, control and the expectations of people who never chose to use the assistant.

Our read

Personalisation is not just remembering that someone likes coffee. The important test is whether users can see, correct and delete what Muse has assembled about people around them, and understand what information it used. Meta describes controls for the user’s data; the harder question is how those controls apply to profiles and inferences about other people.

What to watch

  • Whether users can inspect and edit individual relationship profiles and their sources.
  • How Muse handles incorrect inferences or information about people who do not use the service.
  • Whether Meta explains how long these profiles are retained and how deletion works.

Discussion spark: Should an AI assistant be allowed to build profiles of your friends and family from information you share, or should people be able to opt out even if they never use the assistant?

Sources and evidence

not affiliated with, endorsed by, or operated by Meta