Meta Launches Muse AI Agent to Automate Email, Payments, Tasks
The autonomous assistant can book travel and negotiate bills, but Meta must overcome trust barriers tied to its data practices.
Meta has introduced Muse, a personal AI agent designed to autonomously handle tasks ranging from sending emails to negotiating bills on behalf of users. CEO Mark Zuckerberg announced the launch via Instagram, positioning the agent as a tool that can actively manage digital tasks rather than simply answer questions.
The agent runs on Muse Spark, which Meta describes as its most capable model for agentic work—AI systems that can take actions independently rather than just provide information or recommendations.
How Muse works
Muse is available as a standalone iOS and Android app, through muse.ai, and within WhatsApp. The service is free for basic tasks, with subscription tiers for users who want the agent to handle longer or more complex operations. Meta plans to integrate Muse into its AI glasses in the near future.
The agent accesses users' personal information, web credentials, and payment details to work across multiple applications. It learns from conversations and user preferences over time, according to Meta.
Practical applications include converting saved Instagram recipe videos into grocery lists, suggesting dinner party menus while remembering guests' dietary restrictions, and sending invitations. For more involved tasks—such as negotiating a higher price when selling a car or reducing a monthly bill—Muse continues working in the background after users close the app, returning when it needs approval or when circumstances change.
For purchases, the agent uses Link, an agent wallet built by Stripe, to complete checkout processes.
Why it matters
Meta's move into autonomous AI agents represents a significant escalation in how much control users might cede to automated systems. Unlike chatbots that simply provide recommendations, agents like Muse can take consequential actions—sending messages, making purchases, and accessing sensitive accounts. This raises the stakes considerably for privacy and security, particularly given Meta's history of monetizing user data for advertising and past controversies around user wellbeing. The success of agentic AI may hinge less on technical capability than on whether companies can earn sufficient user trust to access the credentials and permissions these systems require.
The trust challenge
Meta faces substantial hurdles in convincing users to grant an AI agent access to their email, payment information, and online accounts. The company's advertising business model relies on harvesting personal data from social media users, and it has faced criticism over its handling of young users' mental health.
Miranda Bogen, chief technologist at the Center for Democracy and Technology, warned that the risks extend beyond bad recommendations. "With minimal input from users, these systems could take actions that damage someone's reputation, leak sensitive medical information, or even deplete bank accounts," she said in a statement.
Meta has outlined several privacy controls in response to these concerns. The company says Muse does not share user conversations or virtual machine data with Meta's advertising systems. Users can opt out of having their interactions used to train AI models. They can also choose which apps Muse connects to and the level of access granted—for example, whether the agent can only read emails or also send them. Users can disconnect services, change permissions, or instruct Muse to forget specific information at any time.
The details were first reported by Fast Company.
This is an original analysis by the Omega editorial team. Source reporting: AI Watch.
Want systems like this working for your business?
Book a Call
