Meta introduced a new product this week called Muse. The company describes it as the first personal AI agent built for everyone.
Unlike a typical chatbot, Muse does not just answer questions. It carries out tasks on a person's behalf.
People can ask Muse to handle small jobs, like sending an email or booking a trip. They can also give it larger goals, like planning a big event or managing a budget over time.
Muse then builds a plan and starts working. It can open a browser, fill out forms, and even negotiate on a person's behalf.
Meta says the tool can help sell a car for more money or lower a monthly bill. It can also adjust a training plan as someone's schedule changes.
How Muse Works
Muse is powered by Spark, which Meta calls its most capable model built for real-world tasks like this.
The agent keeps working even after a person closes the app. It comes back later if something changes or if it needs approval before finishing a task.
Before it sends an email or makes a purchase, Muse checks with the person first. This step is meant to keep people in control of what the agent does.
Muse also remembers details from past conversations. Meta says it can turn a saved recipe video into a grocery list or suggest a dinner menu based on what a person has said before.
When Muse needs to pay for something, it uses Link, a tool built by Stripe. Meta says Muse is the first AI agent to get purchase protections through Link, including coverage for damaged items and price drops.
Shop Pay is expected to arrive as another payment option. Support for 1Password is also planned, so Muse can use logins people already have saved.
Privacy and Security Features
Muse runs on what Meta calls Muse Secure VM, a dedicated virtual computer built just for this kind of agent.
A separate program called Sentinel operates on the same machine. Nothing Muse does can reach the internet unless Sentinel approves it first.
Meta says Muse cannot see a person's passwords or payment details. Any login information a person shares is stored separately, so the agent can use it without viewing it directly.
People can choose exactly which apps Muse connects to. They can also decide how much access it gets, such as whether it can read email or only send messages.
Access can be changed or removed at any time. People can also choose not to have their conversations used to train Meta's AI models.
Meta plans to release a version called Muse Confidential VM later this year. That version will encrypt a person's data with a key only they control.
Muse is now available in the US on iOS, Android, and muse.ai. Meta says it is coming soon to its AI glasses, and it will be free for most features with paid plans for added use.
