Meta Launches Muse Personal AI Agent
Meta has launched Muse, a personal AI agent designed to handle everyday tasks for users.
Initially, Muse is available to adults in the United States through a dedicated app and WhatsApp. Meta also plans to bring the agent to its smart glasses.
Unlike a conventional chatbot, Muse can take actions on a user’s behalf. For example, it can send emails, book travel, fill out forms and help manage purchases. It can also work on longer-term goals and create personalized plans.
The launch comes as technology companies compete to develop AI agents that can operate with greater independence. However, Meta is placing particular emphasis on security and privacy as Muse gains access to personal services.
Muse Connects With Everyday Apps
Muse runs on Meta’s Muse Secure VM, a dedicated virtual machine with its own browser. The system stores the agent and user data within that environment.
Furthermore, users decide which applications Muse can access. They can also change permissions or disconnect services at any time. For sensitive actions, Muse asks for approval before proceeding.
The agent can connect with services covering email, calendars, payments, shopping, health, smart-home systems and other areas. As a result, Muse can handle tasks that normally require several separate applications.
Muse can also continue working after users close the application. Then, it can return when conditions change or when it needs approval for an action.
For payments, Meta says Muse can use Stripe’s Link service. The system generates a one-time card number, which keeps the user’s actual card details hidden during eligible purchases.
Meta Builds Multiple Safety Controls
Meta has added a separate Sentinel agent to monitor Muse’s actions. Sentinel can block internet access and request user approval when a task requires additional authorization.
Moreover, Muse provides an audit trail showing actions it has taken and plans to take. Users can also opt out of having their interactions used to train Meta’s AI models.
Meta says Muse does not share conversations or data stored inside its virtual machine with its advertising systems. Later this year, the company plans to introduce Muse Confidential VM with user-controlled encryption.
However, the safety claims face scrutiny. Reuters reported that internal testing uncovered security flaws and reliability problems. One reported test involved Muse accessing private iCloud photos after receiving a prompt about objects in birthday-party images.
Meta also delayed Muse’s planned April release while it worked on security improvements. Vishal Shah, Meta’s vice president of AI products, said the company had reached the minimum safety threshold needed for release.
Muse Targets Meta’s Personal AI Ambitions
Meta describes Muse as an early step toward what CEO Mark Zuckerberg calls “personal superintelligence.” The company wants agents to manage more complex tasks while learning users’ preferences over time.
At the same time, Meta faces competition from other AI agent developers. The company is positioning Muse around its large ecosystem of messaging, social, commerce and connected-device services.
For now, Meta offers Muse for free for basic usage, while paid plans target heavier users. Reuters reported subscription prices of $20 and $100 per month.
Nevertheless, trust remains central to the product’s prospects. Muse needs access to sensitive information to do useful work, but that same access creates greater consequences when an agent makes a mistake.
Therefore, Meta’s rollout will test whether users are willing to give an autonomous AI agent meaningful control over their digital lives.
