Muse maps your relationships hourly

Meta's Muse AI assistant, used by four million people, updates detailed dossiers on users and people they mention each hour, recording how relationships formed, shared interests, disputes and tensions within social groups.

Muse profiles people who don't use it

Even non-users are subject to Muse's mapping when other people's agents mention them in chats, messages and emails. Muse is designed to strengthen relationships by noticing inside jokes and shared context.

Muse infers your hidden goals

Muse identifies users' goals including those not said aloud and learns which prompts work best on each person. Internal instructions show examples like this user responds better to short nudges after ten PM.

“This user responds better to short nudges after 10 PM”

Muse shares insights across its network

Although Meta says each user's virtual machine is isolated, Muse agents across many virtual machines teach each other through shared lessons. Meta de-identifies user names before sharing insights about what nudges work best.

“Muse agents across many VMs teach each other through shared lessons”

Meta makes Muse's thinking visible

Users can access Muse's internal files and session logs through the file browser to see how the assistant thinks. Meta Superintelligence Labs' David Singleton said on X the company wants users to explore the system's internals.

Muse hides what it won't erase

Muse's instructions tell it never to tell users that original messages may remain visible in the chat after they ask it to forget something. The instructions explicitly say do not frame that as something Muse failed to erase.

“Do not tell the user that their original messages may remain visible in the chat, and do not frame that as something Muse failed to erase”

Encryption coming but not yet available

Meta has said it plans to offer users encryption later this year that would prevent even the company from accessing Muse data. That option is not yet available to the four million current users.