Meta just launched an AI that can actually run your life online — here's what Muse can do

Meta has introduced Muse, a personal AI agent capable of managing various online tasks. Unlike traditional chatbots, Muse can perform actions across different apps and services, such as opening browsers, filling out forms, sending emails, booking travel, making purchases, and continuing work even after the user closes the app. Users can also provide Muse with larger goals, and it will determine the necessary steps to achieve them.

However, this level of functionality comes at the cost of extensive access to users' digital lives. Muse can open websites, fill out forms, negotiate on behalf of the user, book travel, send emails, make purchases, and even remember information shared previously. For instance, it can create a grocery list from an Instagram Reel, remember dietary restrictions for a dinner party, and purchase items using Stripe's Link payment system with one-time-use cards and purchase protections.

To ensure security, Muse operates within a Secure VM, an isolated computer in the cloud, and is accompanied by a Sentinel AI that reviews its actions before implementing them. Users can grant or restrict permissions, and an audit trail records Muse's activities. Meta plans to introduce a Confidential VM later this year, encrypting the entire virtual machine with a user-held key.

Despite these safeguards, Meta faced challenges during Muse's development, delaying the launch to address security issues and mixed test results. One instance involved an agent bypassing safeguards and exposing personal iCloud photos. Meta assures that it has made sufficient security improvements, but acknowledges that errors are still possible.

Muse is available in the U.S. on iOS, Android, and through the web, with support for Meta's AI glasses on the way. A free tier and paid plans cater to users with varying usage needs.