Meta launches Muse, a personal AI agent for everyday tasks

Meta has introduced Muse, a personal AI agent designed to complete tasks such as organizing information, booking appointments, buying tickets and filling out online forms. The product runs on a dedicated virtual computer in Meta’s cloud and is initially available to adults in the United States through iOS, Android, the web, WhatsApp and, more recently, a Mac app.

Muse is Meta’s latest attempt to turn generative AI from a question-and-answer tool into software that can act on a user’s behalf. The company positions it as a personal assistant that can continue working on longer tasks after a user closes the app. Users can give the agent a name and avatar, set its communication style and connect services including email, calendars, cloud storage and fitness trackers.

The agent is powered by Meta’s Muse Spark 1.3 model, according to Bloomberg. It can connect with services such as Gmail, Google Calendar, Google Drive, Ticketmaster and OpenTable. Meta executive David Singleton says users can also ask Muse to build a connector for a service that offers an application programming interface, or API.

A browser with permissions and approval checks

Muse works through a chat-style interface, but it also has access to a built-in browser within its dedicated virtual machine. Users can see that browser while the agent works. Meta says the separate environment is intended to keep each person’s workspace isolated from others, even though it runs on Meta infrastructure.

The company has built a permissions system called Sentinel to govern what Muse can do online and in connected services. Sentinel can allow an action, block it or send it to the user for approval. Meta says users can distinguish between read access and write access, and can limit permissions to a particular task, service, transaction or time period.

Meta says Muse asks before taking sensitive actions, including making purchases. It can carry out previously approved, lower-risk actions more independently. This balance matters because autonomous agents can cause real damage if they receive overly broad access. Axios notes that companies and users have already reported cases in which AI agents deleted data or attempted automated activity that services did not allow.

Privacy controls remain a central question. Axios reports that Meta can use queries sent to its models by default, though users can turn that option off. The company is also working on a confidential workspace version in which Meta cannot see activity inside the user’s virtual environment.

Free access, subscriptions and a Mac expansion

Meta offers Muse free of charge, with subscription tiers priced at $20 and $100 per month for people who need more computing capacity. Bloomberg reports that the free plan includes up to 100 million tokens each week. Meta executives say they expect most users to remain on the free tier.

There are no advertisements inside Muse. Alexandr Wang, Meta’s chief AI officer, tells Axios that Meta is exploring commerce-related revenue opportunities. The company also plans support for its AI glasses, a move that could bring the agent beyond phones and computers.

On Mac, Muse can organize files, fill out forms and retrieve information from Messages, Calendar and Notes, according to The Verge. The desktop app gives Meta’s agent access to more of the day-to-day work that takes place on a computer, while also increasing the importance of clear permissions and user oversight.

Muse reflects CEO Mark Zuckerberg’s broader ambition to build what Meta calls “personal superintelligence.” Whether consumers trust such agents with enough access to make them genuinely useful will be a key test for the product.

Sources

Stay up to date

AI for content creation: the latest tools, tips and trends. Every two weeks in your inbox:

More info …

About the author

Related posts:

Advertisement

×