Meta launched its personal AI agent, Muse, in the United States on September 8.
How do services that handle real actions—such as email, reservations, and shopping—differ from chatbots that simply provide answers?
More important than early popularity is whether the service is structured so users can entrust it with payments and personal information.
3-line summary
1. Meta Muse is an AI that performs tasks on the user’s behalf
2. U.S. iOS downloads exceeded 73만건
3. Final approval before payment is the key safeguard
A personal AI agent that carries out email and reservations
Meta’s Muse is designed not to stop at answering questions, but to fill out forms in a dedicated browser, send emails, and carry out travel reservations and shopping procedures. Meta explained that tasks requiring a long time can continue to be processed in a dedicated cloud environment even after the app is closed.
The service began rolling out in the United States on iOS, Android, the web, and WhatsApp. Digital Today reported that, as of September 10, U.S. iOS downloads exceeded 8만3000, and that usage on the web and WhatsApp was not included in this tally. Therefore, it is difficult to judge the service’s total user base based solely on its iOS ranking or downloads.
MoneyToday and Newsis reported that Muse reached No. 1 among free apps on the U.S. Apple App Store on September 18 and recorded 73만건 downloads approximately 5 days after launch. An initial response has been confirmed, but this alone does not establish long-term user adoption.
User confirmation is required immediately before payment
When AI handles purchases and payments, malfunctions and permission management become as important as convenience. Meta said it operates the service in the Muse Secure VM, separated for each user, while a separate monitoring AI called Sentinel checks external connections and service manipulation.
In a demonstration introduced by StartupRecipe, a final confirmation screen appears immediately before a purchase becomes an actual order. The order proceeds only after the user reviews the product details and payment amount and grants permission. Meta also explained that it prevents Muse from directly viewing passwords or payment methods.
This is the most important criterion to watch when evaluating Muse. The scope in which AI proposes and prepares actions must be separated from the scope in which it finalizes difficult-to-reverse actions, such as sending emails and making purchases. The value of automation depends not only on how much work it handles, but also on when the user regains control.
The boundary of the service revealed by Amazon’s block
What Muse can do is not determined solely by Meta’s technology. According to MoneyToday and Newsis, Amazon blocked the Muse bot from accessing its website, citing privacy, security, and terms-of-service concerns. If an external service does not permit access, an AI agent’s purchasing and reservation functions may also be limited.
Meta presents Muse as a personal assistant that understands users’ goals and interests. However, the more broadly users entrust it with actions in daily life, the more privacy protection, approval procedures, and cooperation with external services must work together. App rankings are only a starting point; for users, the more direct criteria are what permissions they actually grant and when approval is required.
References
Tags #MetaMuse #Muse #Meta #Muse #AIagent #PersonalAIassistant #Artificialintelligence #AIshopping #AItravelreservations #AIPayments #Privacyprotection #AppStore #WhatsApp