Meta Debuts Muse AI Agent That Runs in Its Own Secure VM

Meta rolled out Muse, a personal AI agent that plugs into a user's email, calendar, payments, health and fitness apps, smart-home devices, dining and shopping services, and completes errands from inside those accounts, TechCrunch reported. It launches for US users at muse.ai, on iOS and Android, and inside WhatsApp, with Meta's AI glasses to follow.

The security pitch is architectural. Muse runs inside its own "dedicated, secure computer with its own browser," the Muse Secure VM, and Meta says the agent "won't have visibility into people's passwords or payment methods." Payments route through Link by Stripe; Shopify Shop Pay and 1Password integrations are listed as coming soon. Meta also says conversations and data are not shared with its ads systems.

The free tier is metered. Paid Power costs $20 a month and Maximum costs $100. The agent is powered by Meta's Muse Spark model.

The bulk of the TechCrunch piece is about trust. The outlet walks through Meta's regulatory history: a 2011 FTC settlement over exposing private information; a 2019 FTC penalty that was "a then record-breaking $5 billion settlement over eight separate privacy-related violations"; a 2023 FTC charge that Meta had violated that same order; an "$18 billion multistate settlement" reached with 29 states in August over social media's consumer harms; and a New Mexico judgment ordering Meta to pay $942 million over harms for children. It also notes the Cambridge Analytica scandal and the 2019 discovery of user passwords stored in readable formats. Muse joins more than a hundred Meta items we have logged in the last ninety days, and it lands into that record rather than around it.