MUMBAI: Meta has introduced Muse, a personal AI agent designed to move beyond answering questions and instead take action on behalf of users.
The company describes Muse as a secure and private AI agent that can help people manage tasks, pursue longer-term goals and proactively suggest ideas. It is being rolled out in the US on iOS, Android and muse.ai, with support for AI glasses planned.
Muse is designed to work through natural conversations, either in its dedicated app or directly through WhatsApp. Users can tell it what they want done, after which the agent can develop a plan, coordinate time and resources and carry out parts of the work itself.
The agent can open a browser, fill in forms and negotiate on a user’s behalf. It can handle everyday tasks such as sending emails and booking travel, while also working on longer-term goals.
For tasks that require more time, Muse can continue working after a user closes the app. It can return when circumstances change or when it requires approval for a sensitive action, such as sending an email or making a purchase.
Meta says Muse can also use information shared in previous conversations to personalise its actions. For example, it can turn a recipe saved on Instagram into a grocery list, suggest a menu for a dinner party and remember guests’ dietary restrictions when preparing invitations.
At the core of Muse is Muse Secure VM, a dedicated virtual machine that houses the agent along with a user’s data and connected service credentials.
Meta says the system is isolated in the cloud, preventing other agents from accessing the environment. A separate Sentinel agent operates at the system level and controls whether Muse can access the internet.
Muse itself does not have visibility into users’ passwords or payment details. Credentials are stored securely so the agent can use connected services without directly seeing the underlying credentials.
The system also requires user approval before sensitive actions and provides an audit trail showing what Muse has done and what it plans to do.
Users can choose which apps Muse connects to and control the level of access granted. They can change permissions or disconnect services at any time. Meta also says users can opt out of having their interactions used to train its AI models.
The company says Muse conversations and data stored in the virtual machine are not shared with Meta’s advertising systems. Users can also instruct Muse to forget specific information it has learned.
Muse can complete purchases using Link, the wallet service built by Stripe. Meta says Muse is the first AI agent covered by Link’s purchase protections, which include coverage for eligible damaged or lost items, price drops, no-fee returns and return guarantees.
Link’s wallet for agents generates a one-time-use card, allowing users to keep their actual card details hidden during online purchases.
Meta said Shop Pay support is coming soon, along with 1Password integration so Muse can use existing logins.
Meta plans to introduce Muse Confidential VM later this year. The company says the system will encrypt the entire virtual machine, including users’ data and conversations, using a key held only by the user.
This would mean Meta itself would not be able to access the contents of the VM.
Muse is powered by Muse Spark, which Meta describes as its most capable model to date and one built specifically for real-world agentic tasks.
Muse is initially rolling out in the US across iOS, Android and the web. Meta said it will also come to AI glasses.
The service will be free for most everyday needs, with subscription plans available for users who want access to additional capabilities.
Meta positioned Muse as an early step towards what it calls personal superintelligence, with the broader aim of shifting AI from a system that primarily responds to requests to one that can independently take on more of the work.