Meta bet on Muse: a personal agent with its own computer
28 Sep 2026 · News
Meta did not ship another chat box. It shipped a personal agent that runs on a dedicated virtual machine, and it still asks before it sends mail or spends money.
On 8 September 2026, Meta introduced Muse: a personal AI agent meant to do work, not only answer questions. Talk to it in the Muse app or in WhatsApp. Tell it a goal. It plans, opens a browser, fills forms, and keeps going after you close the app. The model behind that loop is Muse Spark.
The strange part is the computer. Muse lives on Muse Secure VM, a dedicated cloud virtual machine that holds the agent and the person's data. A separate Sentinel process gates what reaches the internet. Credentials and payment methods stay in secure storage the agent cannot read. Before sensitive steps like sending email or making a purchase, Muse asks the person. There is an audit trail of what it did and what it plans next.
That design matches what we wrote about what a personal agent actually does: a loop with tools and a stop rule. Muse's stop rule is human approval on the moves that hurt if they go wrong.
Meta says Muse is free for most of what people need, with paid plans for heavier use, rolling out in the US on iOS, Android, and muse.ai. Later this year it plans a Confidential VM encrypted with a key only the user holds, so even Meta cannot read the box. Reporting around the launch framed Muse as Meta's big consumer bet to catch up in the agent race.
Simple rule: Meta bet that a personal agent needs its own sealed computer, and a human still owns the send button and the wallet.