Meta launches Muse, an AI agent that does tasks instead of merely chatting
Original publication: · Sergey Zhadov
This is a translated archive article. The date and pricing discussion refer to the original publication, not a new announcement today.
Meta has introduced Muse, a personal AI agent that can open a browser, fill in forms, work on a purchase and continue a longer task after the app closes. That sounds handy. It is not a magic secretary without brakes, though: access currently starts in the United States and sensitive actions require the user's approval.
What Meta actually launched
Muse behaves like a personal agent: you describe a goal, and it breaks the work into steps and acts through its own browser. Meta's examples include email, travel, shopping and longer personal projects. People can talk to it in the Muse app or through WhatsApp.
The important difference from a regular chatbot is action. Muse does not stop after explaining how to buy a ticket. It tries to complete the process. That is useful, but it also raises the stakes. A bad paragraph is annoying. A bad payment is expensive rubbish.
How the agent is supposed to stay under control
Each user gets a separate Muse Secure VM. A distinct Sentinel agent reviews outbound actions. Muse is meant to ask before it sends an email or completes a purchase, and users can inspect a trail of finished and planned steps.
The design sounds sensible, but these are still Meta's product claims. Do not dump your work inbox, payment details and entire digital cupboard into a new service on day one. Start with a harmless chore where a mistake is easy to notice and undo.
Who can use Muse now
The rollout begins in the United States on iOS, Android and the Muse website. Meta says most everyday use will be free, with subscriptions for heavier workloads. AI glasses, Shop Pay, 1Password and a Confidential VM are promised for later.
Later is the crucial word. Those features are not live simply because they appear in an announcement. If Muse is missing in your country, the app is not necessarily broken. Downloading a dodgy APK that promises instant access is a fine way to receive malware instead of an agent.
What changes for ordinary users
Muse shows where the market is heading: chatbots are turning into operators. People will copy fewer answers by hand and approve more concrete actions. That is genuinely convenient when the chain remains visible and every critical operation can be stopped.
The practical conclusion is simple. Let an agent collect options, draft material and handle low-risk routine work. Keep money, publishing, messages sent in your name and data deletion on a short leash. Otherwise automation can turn into one very fast screw-up.
Translation reviewed:
