Meta Launches Muse, an Always-On AI Agent That Runs on Its Own Cloud Computer and Can Email, Book, and Buy for You, Logiciel Tech News

Meta Launches Muse, an Always-On AI Agent That Runs on Its Own Cloud Computer and Can Email, Book, and Buy for You.

Meta launched a personal AI agent called Muse on September 8, and the pitch is that it doesn’t just answer questions, it does the errands. Muse can send messages, book travel, negotiate a bill, turn a recipe video into a grocery list, and buy things on your behalf. It runs on its own dedicated computer in the cloud, with a second AI called Sentinel sitting between it and the open internet, so nothing goes out until Sentinel signs off. Mark Zuckerberg introduced it as “the personal agent that understands your goals and works 24/7 to get things done for you.” It’s live in the US now, free to start, with paid tiers on top.

Robert

Logiciel Tech News

Sep 9, 2026 2 min read

Meta Launches Muse, an Always-On AI Agent That Runs on Its Own Cloud Computer and Can Email, Book, and Buy for You

What Muse Actually Does

The point of Muse is action, not chat. It connects to your email, calendar, payment methods, shopping accounts, and smart home, then works through tasks on its own. Meta’s examples include booking a trip, filling out forms, tracking your preferences over time, and suggesting dinner based on what it knows you like. It runs on a new model Meta calls Muse Spark, which the company describes as its most capable model built for real agentic work.

An always-on agent that acts across your whole digital life is a bigger swing than another chatbot. It’s also the version of this idea most likely to go wrong in public, which is why the rest of the announcement is mostly about safety.

The Security Pitch: A Walled-Off Computer and a Gatekeeper

Muse lives inside what Meta calls the Muse Secure VM, a virtual machine in the cloud that holds both the agent and your data and has its own browser for navigating the web. The Sentinel agent runs on that same machine but is kept separate at the system level. Meta’s line is blunt: “Nothing Muse does reaches the internet unless the Sentinel approves it.” Before anything sensitive, like sending an email or making a purchase, it stops and asks you.

There are real guardrails underneath the marketing. Muse can’t see your passwords or card numbers directly, payments run through Link by Stripe, and Meta says your conversations don’t feed its ad systems. You can pick which apps connect, cut them off anytime, opt out of your data training the model, and tell Muse to forget things. A future “Confidential VM” would encrypt your data with keys only you hold.

The Catch Is Who’s Asking

All of that is a lot of trust to hand one company, and the company is Meta. That’s the tension TechCrunch’s Sarah Perez put front and center, asking plainly whether consumers will trust Meta with this after everything. The record is not gentle: a 2011 FTC settlement over deceiving users, a record $5 billion penalty in 2019 for privacy violations, passwords found stored in plain text, and Cambridge Analytica still hanging over the brand.

So the safeguards may be genuine and still not be the real question. The real question is whether people believe Meta, of all companies, should be the one running an agent that reads their email and moves their money. Good security architecture doesn’t automatically buy back a decade of broken trust.

Price and Where to Get It

Muse is US-only for now, on iOS, Android, and the web, with WhatsApp and Meta’s AI glasses coming later. The free tier gets you limited use. Paid plans run $20 a month for Power and $100 a month for Maximum, per TechCrunch, with connectors to email, calendars, and more layered in as you go up.