Meta Launches Muse AI Agent to Manage Digital Tasks

Meta launched its new personal AI agent, named Muse, on Tuesday, which is designed to actively perform tasks on behalf of users across various digital platforms. CEO Mark Zuckerberg announced the agent in an Instagram post, detailing its capabilities to send emails, book travel, complete web forms, and even negotiate on a user's behalf. The Muse agent operates on Muse Spark, described by Meta as its most capable model specifically engineered for agentic work. The agent is being rolled out in the U.S. as a standalone application for both iOS and Android devices. It is also accessible via the website muse.ai and is integrated into WhatsApp. While most uses of Muse are free, Meta plans to offer subscription tiers for users requiring the agent to handle more extensive or complex tasks. Meta also indicated that its AI glasses will soon incorporate support for Muse. To function, Muse accesses a user's personal information, web credentials, and payment details. The company states that the agent learns from user conversations, identifies what is important to them, and continuously improves its performance over time. An illustrative example provided by Meta shows Muse transforming a saved Instagram recipe reel into a grocery list, suggesting a menu for a dinner party, and remembering guests' dietary restrictions before sending out invitations. For more involved agentic tasks, such as selling a car for a better price or reducing a monthly bill, Muse is designed to continue working even after the user closes the application, returning for approval or when updates occur. For purchasing transactions, Muse can utilize an agent wallet named Link, which is built by Stripe, to complete payments. A significant challenge for the widespread adoption of Muse is user trust, given Meta's business model, which relies on collecting personal data from its social media users, and past criticisms regarding its impact on young users' mental health. The increasing autonomy and agency granted to AI agents like Muse also heighten concerns about potential privacy and security breaches. Miranda Bo, chief technologist at the Center for Democracy and Technology, highlighted that the repercussions of AI agents making errors extend beyond mere poor recommendations or intrusive advertising, suggesting more profound consequences.