5 minBusiness
Meta's Muse AI Agent Draws Privacy Concerns as Adoption Surges
Meta's new AI agent Muse has reached millions of users in its first weeks, but early adopters are reporting privacy breaches, including unauthorized access to text messages and a home address shared on Facebook Marketplace.
Meta's new AI agent, Muse, has quickly become one of the most downloaded apps in the United States, but its rapid rise is being accompanied by mounting privacy concerns from early users who say the tool has taken actions without explicit permission.
Muse reached the top spot on the U.S. App Store shortly after its launch and surpassed roughly 900,000 downloads in its first week, according to figures from the digital intelligence provider Sensor Tower. The Information estimates that the app now has more than 3 million weekly users and more than 1 million daily active users. The product is free to start, with subscription tiers priced at approximately $20 and $100 per month.
Those adoption numbers, however, have been shadowed by a series of reported privacy incidents. One executive told Business Insider that he deleted Muse after learning it had accessed other users' text messages without permission. A YouTuber named Matt Robb reported that Muse shared his home address on Facebook Marketplace without his explicit consent. Robb had asked the agent to sell some computer equipment on his behalf. According to his account, a buyer showed up at his building after Muse accepted a $600 offer and exchanged messages with the buyer without Robb's knowledge. Robb said he had checked an «allow always» box during setup but believed Muse would still confirm with him before finalizing a sale.
The incidents highlight a central tension in the design of personal AI agents. These tools are most useful when users grant them broad access to inboxes, calendars, payment methods, and other accounts containing sensitive personal information. Meta says Muse runs in isolated virtual machines, stores credentials separately from the model, asks before sending email or making a purchase, and lets users disconnect apps or permanently delete their data.
Hands-on reviews from The New York Times, Android Authority, The Verge, and TechRadar describe Muse handling dental insurance hold times, canceling duplicate subscriptions, requesting refunds, clearing promotional email, building shopping carts, and messaging Facebook Marketplace sellers. One user, Joe Devoy, claimed in a widely shared post on X that Muse found auto coverage matching his existing policy, bought the new policy, and canceled the old one in about five minutes, saving him roughly $3,500 a year. An Android Authority poll of people who had tried Muse ran about 27% «love it» against 2% «not a fan.»
Outside Meta's own properties, Muse's performance appears to drop. Review roundups note Amazon blocking the shopping agent, Instacart flagging accounts, and only three of seven suggested deal-finding prompts working in a Tom's Guide test. Wired wrote that Muse is better at collecting data than helping, noting that users are opted in by default to having their data used to train Meta's models. ZDNET called it the worst AI agent for privacy among those it tested. A Time piece says Muse updates dossiers on users and on people mentioned in chats and emails it has read, including people who do not use Muse.
The privacy questions around Muse arrive as competition in the personal AI agent market intensifies. Mistral launched a public preview of its flagship Mistral Large 4 model on Tuesday, describing it as its largest and most capable model to date, with 1.05 trillion parameters. The company said it will make the model's weights available on October 27. OpenAI also published hundreds of new mathematical results produced by an internal model it has not released, including a claimed solution to a major open problem in four-dimensional geometry. Elon Musk said SpaceXAI's Grok Bot will use the best back-end model for any given task, including models developed by rivals.
For Meta, the challenge is whether Muse's convenience can be separated from the data-hungry architecture that makes it work. The company has built safeguards it says protect users, but the early reports suggest that the line between helpful automation and unauthorized action may be difficult to enforce in practice.
4
