Meta launches Muse, an AI agent that can send emails, book trips and make payments

Meta has launched Muse, a new AI assistant designed to do more than answer questions.

The agent can connect to external apps and services to perform tasks such as sending emails, managing calendars, booking travel and making payments. Meta is initially making Muse available in the United States through a dedicated app and WhatsApp.

Internally known as Hatch during development, Muse runs each user instance inside its own virtual machine. That allows the agent to continue handling tasks in the background rather than requiring the user to supervise every action step by step.

The product reflects a broader shift in consumer AI. The major AI companies are moving from assistants that generate text toward agents that can operate software and complete real-world workflows.

Muse can access services including email, calendars, smart-home systems and other connected data when users grant permission. Meta says those permissions can be revoked.

The rollout also raises obvious security and privacy questions. Reuters reported that internal testers encountered problems including unexpected data uploads, service disconnections and cases where sensitive personal information could be exposed.

Meta says the product meets its security and privacy standards while acknowledging that an autonomous agent can still make mistakes.

The important change is not the chatbot interface. It is the amount of authority the software is being given.

AI assistants are increasingly moving from suggesting what a user should do to actually doing it.