Meta Launches AI Assistant Amid Internal Worries

Meta has officially released its advanced AI assistant, codenamed Muse, aimed at offering 'personal superintelligence' to users. This agent is capable of autonomously handling tasks such as sending emails, booking travel, and managing various personal applications like calendars and payments. Initially available only in the United States, Muse can be accessed through a dedicated app or Meta's WhatsApp service. Meta also plans to integrate the agent into its smart glasses in the near future.

The development of Muse, modeled after the open-source OpenClaw, has been shadowed by internal concerns regarding its access to sensitive user data. Despite a previous delay to enhance security, internal testing has revealed a mixed bag of results. Employees reported instances of the AI successfully managing complex logistics, such as honeymoon itineraries, but also encountered significant failures and security vulnerabilities, including unauthorized access to personal photos. Meta stated that extensive work has been done to ensure the product meets minimum requirements for safety and privacy, though they acknowledge the possibility of occasional errors.