Meta’s personal assistant Muse has become a viral hit, with millions of users downloading the AI agent and linking it to bank accounts, messages and health data so it can complete tasks on their behalf. As the app spreads, researchers have extracted internal files and operating instructions that show how the system organizes information about the people in a user’s life. Meta says the files were meant to be accessible as a matter of transparency.
Independent AI safety and security researcher Karan Joshi obtained an extensive set of Muse instructions and system prompts by using the regular chat interface to ask the agent to copy and share its own software files. He shared the findings with WIRED. According to those instructions, Muse can build “a page for every person in the user’s life,” an hourly process that compiles data on family, partners, friends, colleagues, collaborators and people the user follows.
The documentation says a page may begin sparse and fill in over time, with sections such as Facts, History, The relationship, In common, Open threads and Strengthening. Muse is instructed to rely only on available evidence and to treat invented details as worse than an empty page. Pages could record where a person lives, recurring threads such as an apartment move or a shared savings goal, and dates that matter, including birthdays and anniversaries.
Joshi said the prompts and system data suggest Meta wants to understand users’ relationships with real people, and that the effort to know users like a friend is concerning. Carissa Véliz, an associate professor at Oxford’s Institute for Ethics in AI, said AI systems are given far more information about users than users receive about them, including what the systems infer, correctly or incorrectly, and what they piece together from other data sources.
Muse runs each user on a dedicated virtual machine that stores user data and context, according to the report. That machine is inaccessible to other agents, and users can wipe memories or disconnect external services at any time. Meta says Muse is designed to seek human confirmation before actions such as sending an email or making a purchase, and it maintains an audit log showing the agent’s activity and future plans.
Meta spokesperson Daniel Roberts said an agent needs context about a user and the people they interact with to be useful, and that Muse gathers it from public information and what users choose to share. Miranda Bogen, director of the Center for Democracy and Technology’s AI Governance Lab, said Muse appears to place more emphasis on relationships and personal contacts than rival systems, and that such assistants push users to share more of their lives.
More AI news from TechManNews.
