Now loading...
Meta has unveiled Muse, a new personal AI agent designed to assist users in achieving their objectives while ensuring privacy and security. The AI operates on the Muse Secure VM, a specially crafted virtual machine that securely manages both the agent’s operations and user data. Users can communicate with Muse just as they would with a friend, using either the Muse app or WhatsApp.
The application streamlines task management by allowing users to inform Muse of their needs, which then takes action using Muse Spark, Meta’s most advanced AI model optimized for practical applications. Muse is designed for global accessibility, requiring no technical knowledge for users. It can manage a variety of tasks, from sending emails to planning travel, and is capable of tackling larger projects. By sharing goals with Muse, users receive personalized plans and resource coordination, with the agent autonomously advancing tasks like browsing and completing forms.
Muse is also designed to continue work after users exit the app, notifying them of updates or requiring approval before key actions, such as sending messages or making purchases. Its capabilities offer efficient results, including increased sale prices for items or optimized plans for personal training as circumstances evolve.
The payment process is facilitated through Link built by Stripe, making Muse the first AI agent equipped with Link’s purchase protections, which include coverage for lost or damaged items, refunds for price drops, and guarantees on eligible returns. With a one-time-use card provided for online transactions, Muse ensures users’ financial information remains secure. Additionally, integrations with Shop Pay and 1Password will soon be incorporated, enhancing user convenience.
Muse is also equipped to recall important user information, providing unsolicited suggestions based on previous conversations. For example, it can convert a saved recipe from Instagram into a grocery list or remind users of dietary restrictions among friends when organizing gatherings.
Privacy is a top priority for Meta, which has implemented robust security measures within the Muse Secure VM. This dedicated cloud computer ensures operational isolation, with a separate Sentinel agent validating actions before they access the internet. Muse does not view or store sensitive information such as passwords directly, relying instead on secure, managed access to user credentials.
Users maintain control over which applications Muse can link to and what permissions it has, with the ability to revoke access at any time. Personal conversations and data remain confidential, never shared with Meta’s advertising systems, and users can prompt Muse to forget specific details if desired.
Later in the year, Meta plans to improve Muse with the Muse Confidential VM, where all user data and conversations will be fully encrypted, accessible only by the user. Meta envisions that personal superintelligence will revolutionize technology, and considers Muse a significant step towards that goal, enabling users to focus on what truly matters to them.
Muse is currently being rolled out in the U.S. on iOS, Android, and muse.ai, with further availability for AI glasses on the horizon. The platform will be free for standard use, with subscription options for enhanced features.
