Meta, developer of Facebook and Instagram, has launched Muse, a personal AI agent that can work across websites and connected services, continue tasks in the background and interact with users directly through WhatsApp.
Muse enters a growing market for AI agents that can take actions rather than only respond to prompts. Meta is distinguishing its approach with a dedicated virtual computer called Muse Secure VM, along with controls designed to manage credentials, permissions and sensitive actions.
Muse launched on September 8 and has since climbed to the top of the U.S. App Store’s free iPhone rankings, according to Axios. App Store rankings can change frequently, but the early position provides some indication of consumer interest in the product.
Muse is designed to handle multi-step tasks that may involve several websites or services. Users can ask it to book travel, fill out forms, manage purchases or work with connected accounts. Meta says Muse can continue working after the app is closed and return when it needs additional information or approval.
The agent can also help with longer-term goals by creating plans, organizing tasks and continuing work over time.
Muse is powered by Muse Spark, a Meta model designed for agent-based tasks.
WhatsApp Is Part of Meta’s Approach
Muse can be accessed through its own app or directly through WhatsApp.
That gives Meta a way to make the agent available through an existing messaging platform rather than requiring users to interact with it only through a separate AI application.
Meta also plans to bring Muse to its AI glasses.
Muse supports purchases through Stripe’s Link payment system, which can generate a one-time payment card instead of exposing a user’s underlying card details directly to the agent.
Meta says Muse asks for approval before sensitive actions such as sending an email or completing a purchase.
Support for Shopify’s Shop Pay and 1Password is also planned, allowing Muse to work with additional payment methods and saved login credentials.
Because Muse is designed to interact with outside services, users can choose which accounts and information it is allowed to access.
Runs Inside a Dedicated Virtual Computer
Each Muse operates inside a Muse Secure VM, a dedicated virtual computer in the cloud that contains the agent, its browser and data associated with connected services.
Meta says credentials placed in secure storage can be used by Muse without the agent directly seeing passwords or payment information.
A separate system called Sentinel monitors actions leaving the virtual machine and can require user approval before certain actions are completed.
Users can also review an audit trail showing what Muse has done and what it plans to do. Meta says people can change permissions or disconnect services at any time.
The company also says conversations and information stored inside a user’s Muse virtual machine are not shared with Meta’s advertising systems.
Meta plans to introduce a Muse Confidential VM later this year, which it says will encrypt the virtual machine and its data using a key controlled by the user.
Muse is available on iOS, Android and the web, with Meta also planning to bring the agent to its AI glasses. Meta says most functionality is free, with subscription plans available for additional usage.
Meta plans to expand Muse with support for Shop Pay and 1Password, along with availability on its AI glasses. Later this year, the company also plans to introduce Muse Confidential VM, a more private version of its virtual environment designed to keep a user’s data encrypted.
Leave a Reply