Article Complete
AIM
Meta Launches Muse AI Agent With Internet Access Guardrails
Meta Launches Muse AI Agent With Internet Access Guardrails

Meta Launches Muse AI Agent With Internet Access Guardrails

Meta says Muse uses a dedicated virtual machine and a separate Sentinel system to control internet access, credentials, and sensitive actions.

Published Updated

Key takeaways

  • Meta has launched Muse, an AI agent that assists users with tasks like sending emails and booking travel.
  • Muse operates on a dedicated virtual machine with a separate system, Sentinel, controlling its internet access and sensitive actions.
  • The service is free for most users, with subscription plans available for additional features.
  • User approval is required for sensitive actions, and an audit trail is provided for transparency.
  • Muse can retain and use information from past interactions, but users can request it to forget specific details.
5 min

Meta has introduced Muse, a personal AI agent that can perform tasks on users’ behalf, including sending emails, booking travel, filling out forms, and making purchases.

Muse is being rolled out in the US on iOS, Android, and muse.ai, with support for AI glasses planned. Meta said the service is free for most use cases, while subscription plans will be available for users who want additional capabilities.

Muse is designed to work after a user closes the app, and can develop plans around longer-term goals, coordinate resources, and take action when circumstances change. The agent can also return to the user when it needs approval for actions such as sending an email or completing a purchase.

“People just tell Muse what needs to get done, and it takes action,” Meta said in a statement.

Muse runs on Muse Secure VM, a dedicated virtual machine that contains the agent and a user’s data. It has its own browser and can interact with services that a user connects to it.

Meta said a separate Sentinel agent operates on the same machine and controls Muse’s access to the internet. “Nothing Muse does reaches the internet unless the Sentinel approves it,” the company said.

The system keeps credentials hidden from Muse. Passwords and payment information are stored in secure storage, allowing the agent to use them without accessing the underlying credentials.

Muse also requires user approval for sensitive actions and provides an audit trail showing what it has done and what it plans to do. Users can decide which services Muse can access and set permissions, including whether it can only read email or also send messages on their behalf.

The agent can retain information from conversations and use it in future tasks. Meta said users can ask Muse to forget specific information and can opt out of having their interactions used to train Meta’s AI models.

For purchases, Muse can use Link, Stripe’s payment service, through a wallet for AI agents that generates a one-time-use card. Meta said Muse is the first AI agent covered by Link’s purchase protections. Support for Shop Pay and 1Password is planned.

Meta also said it plans to introduce Muse Confidential VM later this year. Under the system, the entire virtual machine, including user data and conversations, will be encrypted with a key held only by the user.

Muse is powered by Muse Spark, which Meta described as its most capable model to date and designed for agentic tasks.



Membership

Stay ahead with AI insights.

Unlimited access to premium articles, exclusive analysis, and the stories shaping India's AI economy.

Subscribe now

AIM Americas

1190 Miraloma Way Suite P, Sunnyvale, CA 94085

AIM India

1st Floor, Sakti Statesman, Bellandur, Bengaluru – 560103

© 2026 AIM. All rights reserved.