Meta launches "Muse," a personal AI agent that can email, shop, and book travel for you

Meta wants its AI to stop suggesting and start doing. The company launched Muse on Tuesday, a personal AI agent that can send emails, book travel, sell a car and make payments on a user's behalf, rolling out to U.S. adults through a dedicated app, the web, and WhatsApp, with support for Meta's smart glasses planned.

Muse runs on Muse Spark, Meta's agentic AI model, and each user gets a dedicated cloud machine called Muse Secure VM that isolates their agent, data, and connected-service credentials from everyone else's. A separate system called Sentinel reviews actions before they reach the internet. A few details define how it actually works:

• Purchases run through Link by Stripe, generating single-use card numbers so real payment details stay hidden from the agent
• Muse connects to categories including email, calendars, payments, health, shopping, and smart home apps, with users choosing exactly what it can access
• Pricing spans a free tier up through $20 and $100 monthly subscriptions for heavier use

But the launch arrives with a complication Meta didn't put in its press release. According to Reuters reporting cited by Forbes, Meta employees testing Muse in the days before launch flagged security failures and unreliable behavior, including instances of private data exposure.

The timing compounds the pressure: Muse arrives just two weeks after Meta agreed to a settlement of up to $17.1 billion over claims it designed Instagram and Facebook to be addictive to children, meaning an agent capable of touching a user's inbox, wallet, and health data is launching into a moment when Meta's handling of personal information is already under an unusually bright spotlight.