On September 9, Meta Platforms (META.O) launched its personal AI agent application, Muse. Users can entrust digital tasks such as scheduling appointments, filling out electronic forms, and monitoring home cameras to the AI assistant through this app. According to Alexander Wang, head of Meta AI, Muse is powered by Meta's Muse Spark series of foundational models and offers a free version along with two paid subscription tiers at $20 and $100 per month. He emphasized that Muse operates in an isolated environment within Meta's computational infrastructure, does not access users' actual passwords or payment information, and requests user confirmation before performing sensitive operations. Users can choose whether to allow Meta to use their interactions with Muse to train AI models. If permitted, Meta will use the data for model improvement after removing key personally identifiable information. This launch comes at a time when Meta is facing pressures related to AI safety, privacy concerns, and the return on investment in AI. Meta CEO Mark Zuckerberg has previously stated that personal AI agents are a crucial direction for the company's future product and revenue growth. Additionally, Meta revealed that it will explore business models related to shopping transactions with AI agents in the future.