Social media (SNS) corporations Meta unveiled a personal artificial intelligence (AI) agent that performs tasks on behalf of users.
Meta said on the 8th (local time) that it will launch Muse, an AI agent that analyzes a user's goals, devises plans on its own, and performs tasks. Meta introduced Muse as the world's first personal agent focused on personal errands, unlike existing agents.
When instructed by a user, Muse writes and sends emails via a dedicated app, makes travel-related reservations, and completes shopping. Even if the smartphone app is closed, the work continues in the background, and it sends a notification only when a decision is needed, such as final email sending or payment, to request approval.
In particular, it said that through a partnership with the financial platform Stripe, it generates a one-time virtual card number for shopping payments to block card information leaks at the source. It added that there would be no issue applying compensation if a problem arises with a product in this process.
Muse remembers a user's personal information and past conversation history and uses it to respond. To protect personal information, it runs in a segregated environment, the "Muse secure virtual machine" (secure VM). Meta said it will not share conversations with Muse or data within the virtual machine with its own advertising system.
At the end of this year, it plans to operate the "Muse confidential virtual machine" (confidential VM), a fully encrypted technology, so that no one other than the user can access data within the virtual machine. Muse is currently available in the United States on smartphones and through its own webpage, muse.ai. It will also be introduced for AI glasses.
Mark Zuckerberg, Meta chief executive officer (CEO), said in an interview with the tech outlet Source, "The business model we anticipate is to take only a very small percentage as a fee from the commerce amount (handled by the AI agent)," adding, "The fees will come from corporations, not individuals."
However, Reuters, citing Meta's internal test materials, reported that errors were found, including Muse exposing an entire personal photo album, repeated logouts, and disconnections.
Vishal Shah, Meta's vice president in charge of AI products, said, "We cannot assert that (the AI agent) will never make mistakes, but every part of the architecture is designed to be as safe, secure, and privacy-protective as possible."