Meta on Tuesday, September 8, introduced Muse, a personal AI agent that can do more than simply answer questions. It can independently perform tasks across third-party services, including sending emails, booking trips, purchasing goods and making payments.
Muse is part of Meta’s strategy to build what CEO Mark Zuckerberg calls “personal superintelligence.”
What Muse can do
- Send emails, book trips, sell products and make payments on the user’s behalf.
- Handle long-term tasks. For example, it can create a year-long workout plan or help organize the launch of a new business.
- Break a goal down into individual steps, plan how to complete them and continue working without constant input from the user.
- Interact with the web browser by opening websites, filling out forms and negotiating on the user’s behalf.
- Access email, calendars, payment services, shopping apps, healthcare services and smart home systems.
The user decides which apps and services to connect, and access can be revoked at any time.
Availability and pricing
Muse is currently available only in the United States and only to users aged 18 and over.
The service can be accessed through a standalone app, the muse.ai website, iOS and Android apps, as well as WhatsApp.
Meta also plans to integrate Muse into its smart glasses, although the company has not announced a timeline for the rollout.
Basic use is free. Users who need higher limits can choose between two paid plans:
- Power — $20 per month
- Maximum — $100 per month
Users must link a payment card to get started. The app also displays the remaining free usage allowance.
Security
The launch of Muse was delayed from April. According to Vishal Shah, Meta’s vice president of AI products, the company used the additional time to improve the system’s security.
Muse runs on Meta’s own model, Muse Spark. The agent operates inside a separate secure virtual machine called Muse Secure VM, where both the agent and user data are stored.
Why the launch of Muse raises questions
Meta is bringing Muse to market amid a series of major legal battles.
Less than two weeks before the launch, the company agreed to pay $18 billion as part of a multistate settlement over claims related to the harm caused by social media to users. In August, Meta also settled a lawsuit involving 29 states.
A separate case in New Mexico, concerning harm to children, is also ongoing. In that case, a court ordered the company to pay $942 million.
Against this backdrop, Muse raises an obvious question: how much access to personal information will users be willing to give Meta? The agent may require access to email, calendars, payments, purchases and other personal data to perform tasks on their behalf.
Trust will be one of the biggest challenges for Meta’s new product. The company is asking users to do more than simply chat with AI — it wants them to delegate parts of their everyday lives and financial transactions to an autonomous agent.
AI agents are emerging as the next step beyond conventional chatbots. Instead of simply responding to a request, they are designed to take action and complete tasks from start to finish.
