U.S. — Meta launched the Muse AI assistant app in the U.S. on September 9, 2026. The application offers a personal AI agent with autonomous capabilities across iOS, Android, web, WhatsApp, and Mac platforms.

The app allows users to delegate tasks such as sending emails, booking travel, managing schedules, shopping on their behalf, and placing calls to U.S. businesses. To execute these functions, Muse uses a virtual machine to browse the internet on a user's behalf. The system operates on a dedicated computer in the cloud that is isolated so that another user’s AI agent cannot access it.

On the Mac platform, the app can interact with files, messages, calendar, notes, and mail within their native applications. Mark Zuckerberg stated on X regarding the Mac app launch, "The team is shipping fast." David Singleton of Meta Superintelligence Labs stated that Muse "does not watch notifications on your Mac, but rather syncs data from Messages only after the user has specifically enabled access," and apologized for an incorrect response from Muse regarding how it accesses message data.

The product saw rapid adoption following its release. Muse has been downloaded over 900,000 times in its first week, according to reports. The Muse app reached the number 2 position in the U.S. App Store top charts at one point. The service is available with a free tier and $20 and $100 monthly subscription options.

Muse stores details about user interactions and preferences in a Memory document. Users can send a chat request for Muse to wipe memory data and manually edit the Memory file, but Meta does not currently offer a toggle to turn off the memory feature altogether. Users can request and download their Muse agent data by selecting Settings, scrolling to Data Controls, and selecting the option to download agent data.

Muse users are automatically opted in to having their interactions with the agent used for AI model training. Users can turn off AI model training data collection in the Muse app settings under Data controls by disabling the toggle labeled Help improve our AI models. Meta says interaction data used for training is sanitized to remove identifying information. Tarek Sheasha, vice president at Meta Superintelligence Labs, stated in a blog post, "We think this is a good default—every Muse user gets a better personal agent as we all collectively use the product and help the model understand the intricacies of human life."

Meta delayed the Muse launch from April 2026 to September 2026 to improve security and added an autonomous safety agent that monitors Muse's actions. Internal testing uncovered an incident in which Muse exposed private iCloud photos. Meta employees reported repeated logouts, monitoring failures, and inconsistent performance during Muse testing. Meta plans to release confidential versions of its virtual machine later in 2026 that cryptographically prevent the company from accessing the data inside.

Rory Mir, director of open access at the Electronic Frontier Foundation, stated, "Folks don’t recognize that when you talk to an AI, you are talking to the company hosting the AI. These chat windows—that we’re used to being connections between us and another person—are really just us directly putting information into Meta servers about ourselves." Calli Schroeder, senior counsel at the Electronic Privacy Information Center, stated, "This tells me they have not learned from past mistakes and undermines their argument that you should trust Muse with all your information even further." Eric Wong, an assistant professor in the Department of Computer Science at the University of Pennsylvania, stated, "It is true that these companies are trying to push more into personal assistance. You really do have to be careful about what you tell it to do and what information you give it. Engineers make mistakes sometimes. Not all apps are perfectly programmed. Because everything is so new, the products come out, they get adapted, and if something goes wrong, it’s up to the consumer right now." Meta spokesperson Emil Vazquez stated, "Muse is the first personal AI agent built for everyone, with built-in protections and user controls that put people absolutely in charge of how they use it—any suggestion we didn’t build with that in mind from the beginning is ludicrous."

Why It Matters

The launch introduces an autonomous agent capable of executing complex digital tasks, raising questions about data privacy and security in personal AI assistants. The automatic opt-in for model training and the lack of a complete memory toggle reflect tensions between product improvement and user control over personal data.

Internal testing issues, including the exposure of private iCloud photos and performance inconsistencies, occurred despite a five-month delay intended to improve security. The company's plan to release cryptographically secure virtual machines later in 2026 indicates ongoing efforts to address data access concerns while scaling the service.