Meta's Muse Is Better at Surveilling Than Helping Me

Meta's new AI agent automatically opts users into AI training, which can be turned off but requires manual action.

Meta has launched Muse, a personal AI agent designed to handle everyday tasks such as managing inboxes, finding deals, and booking reservations. The app, which is free to use and connects to user data sources like email and bank accounts, has seen over 900,000 downloads in its first week. Muse operates by using a "virtual machine" to browse the internet on behalf of the user, mimicking human browsing and task execution with a high degree of accuracy. Users can interact with Muse through its dedicated app or Meta's WhatsApp platform. While Meta claims Muse has built-in protections and user controls, consumer advocates express concerns that the tool is more focused on data collection than user assistance, with users automatically opted into AI model training unless they manually disable it. Meta spokesperson Emil Vazquez stated that any suggestion Muse wasn't built with user control in mind is "ludicrous."

Meta's Muse launch signals a significant push to embed AI agents directly into daily consumer life, leveraging its vast social media ecosystem for user acquisition and data integration. The app's default opt-in for AI model training, despite user controls, raises immediate privacy concerns and echoes past controversies, potentially impacting user trust and industry standards for AI data handling. The tool's extensive data access, from email to financial accounts, positions it as a potent data collection engine, which could indirectly influence ad targeting and future revenue streams beyond advertising, as hinted by Meta's leadership.

The operational model of Muse, using virtual machines for web browsing, represents an advancement in AI agent capabilities, promising more seamless task automation. However, this sophistication also heightens concerns about potential cognitive degradation and user disengagement, as individuals may cede decision-making and personal taste development to the AI. Consumer advocates warn that such reliance could lead to a passive user base, overly dependent on automated recommendations and potentially susceptible to brand prioritization by AI systems, further entrenching Meta's influence over user perception and behavior.