Meta Launches Muse AI Agent for Autonomous Tasks

Meta introduced Muse on Tuesday, September 8, a personal AI agent engineered to autonomously execute everyday tasks for users in the U.S. This launch signifies a shift from passive chatbots to proactive agents designed to act on a user's behalf, requiring deep integration into daily digital routines. Muse Spark, the system's core, connects to everyday apps and services including email, calendars, payments, health and fitness trackers, smart home devices, dining platforms, and shopping sites. Users control this integration by opting into each connection individually. Once connected, Muse can perform a variety of tasks such as sending emails, booking travel arrangements, negotiating lower bills, completing forms, creating plans, and making purchases. For secure transactions, Muse incorporates checkout protections via Link by Stripe, with planned integrations for Shopify and 1Password. Users can personalize Muse by selecting an avatar, assigning a name, and adjusting communication preferences. The agent's technical infrastructure supports advanced autonomy. If a linked service offers a public API, Muse establishes a connection using user-provided credentials. For services lacking an API, Muse interacts through a web browser. Tarek Sheasha, Meta Superintelligence Labs VP, stated that the AI model was trained for zero-shot tool calling, utilizing command-line interfaces (CLIs) and specific skills. Muse is capable of operating in the background, even after the user exits an application. The system can autonomously deploy swarms of subagents, develop its own tools, and modify its own code to enhance performance. Meta has implemented rigorous digital safety measures to govern Muse's operations. The service is available in the U.S. with a free tier and two premium subscription options, all operating within an isolated, secure virtual machine environment. This development positions Meta to compete in the emerging market of AI agents capable of complex, real-world task execution, raising considerations about data privacy and the extent of AI autonomy in personal digital ecosystems.