Meta develops Muse AI agent to execute real-world tasks

Meta is developing Muse, an AI agent designed to transition from a conversational chatbot to an active 'executor' of tasks. According to Meta AI lead Alexandr Wang, the goal is to reduce the friction between human intention and real-world action by handling administrative tasks, scheduling, and communication.

Unlike traditional chatbots that primarily provide information, Muse aims to interpret user intent and execute multi-step workflows, such as sending emails, managing calendars, and coordinating with external services. Meta has announced integrations with platforms including Shopify, PayPal, Expedia, Instacart, Notion, GitHub, and Box. The technology is expected to be integrated into AI glasses to allow for hands-free interaction.

However, the rollout faces technical and security challenges. Reports indicate a critical security vulnerability that could allow unauthorized access to user accounts and files via voice commands. Additionally, the massive computational requirements for scaling Muse to 100 million users—estimated at 200 million vCPUs and 10,000 petabytes of SSD capacity—pose significant infrastructure hurdles.