Meta unveils Muse AI assistant that can operate across apps and handle tasks
Meta released Muse, an AI agent that can access users' email, calendar, payments and other apps to perform actions such as sending messages and booking travel, with a free tier and paid subscriptions.
Meta introduced Muse, an artificial-intelligence assistant that can interact with a user’s apps—including email, calendar, payment services and smart-home devices—to carry out tasks such as sending messages, making purchases and arranging travel. The service is initially available only in the United States through a dedicated Muse app and via WhatsApp, with plans to add it to the company’s smart-glasses lineup later.
A basic version will be offered for free, while subscription plans cost $20 per month and $100 per month for higher usage, according to a company spokesperson. Users can select which apps the agent can access and revoke permissions at any time. Each Muse instance runs on its own cloud-based virtual machine, allowing it to continue processing requests even when the user is not actively engaged.
Meta said the system includes safety features, such as a monitoring component that can request user authorization before executing certain actions. Vishal Shah, Meta’s vice president of AI products, told Reuters that the rollout was delayed in April to address security concerns and that the architecture has been designed to be “as safe, as secure, as private as we can possibly make it.”
Internal testing reports shared with Reuters showed mixed results. Some employees praised Muse for handling vacation planning, while others reported security lapses, such as the agent exposing personal photos when asked to identify objects in images, and reliability issues like frequent log-outs and failure to refresh pages. Meta did not comment on those specific incidents.
The launch occurs as other AI developers have faced similar challenges with agents that sometimes bypass safeguards or behave unpredictably, and Meta has noted a rise in technical and security incidents linked to AI-driven tools.