Meta has unveiled its new autonomous personal AI assistant, Muse (internal codenamed Hatch), which can perform tasks on behalf of users in other apps, such as sending emails, selling cars, or booking trips.
Meta has unveiled its new autonomous personal AI assistant, Muse (internal codenamed Hatch), which can perform tasks on behalf of users in other apps, such as sending emails, selling cars, or booking trips.
The new product is a key part of CEO Mark Zuckerberg’s strategy to bring “personal superintelligence” to billions of users. It’s also part of Meta’s plan to diversify revenue sources beyond advertising and recoup massive investments in AI infrastructure, which are expected to top $130 billion this year, Reuters reports .
Built on the open source AI agent OpenClaw, Muse can connect to applications in email, calendaring, payments, health, shopping, and smart home. Users choose which services are integrated and can revoke access at any time.
Each agent runs on its own virtual machine in the Meta cloud, allowing Muse to continue performing tasks even when the user has closed the app or is not interacting with the device.
The basic version of Muse is available for free. For users with more advanced needs, there are paid plans — $20 and $100 per month.
The launch has begun in the US. The application is available through a separate ET CIO platform and on WhatsApp. The company plans to add support for Ray-Ban Meta smart glasses soon.
Synchronization with real personal data significantly increases privacy risks. Meta’s VP of AI products, Vishal Shah, noted that the release was intentionally delayed from April to strengthen security and reach the required minimum threshold. In particular, the system built a separate AI controller that tracks Muse’s actions and requires user authorization for sensitive operations.
However, internal testing by Meta employees before launch showed mixed results. One tester reported that the agent bypassed security barriers and exposed private iCloud photos while trying to identify toys in photos from a children's party.
Meta CTO Andrew Bosworth complained about being constantly logged out of the system. Another employee noted that the agent became unreliable when monitoring tickets, stopping to refresh the page after 15 minutes and ignoring errors.
Meta has not yet provided any official comments regarding the failures during internal testing.
