Meta Launches Muse AI Agent to Run Errands, Book Travel, and Shop

Meta has unveiled Muse, an AI personal agent capable of carrying out tasks such as booking travel, making purchases, sending emails, and handling other requests. This innovative tool, initially launched in the United States for iOS, Android, web, and WhatsApp users, represents a significant step up from traditional chatbots by enabling actual action rather than just conversational exchanges. Initially available for free, Meta also introduced two paid plans: Power at $20 per month and Maximum at $100 per month.

Muse leverages Muse Spark, Meta's latest AI model, to function continuously even after users close the app. It goes beyond standard assistants by possessing access to a browser, file system, and terminal, allowing it to interact with websites, complete forms, and construct tools as needed. Users have the ability to customize their Muse by assigning a name, selecting an avatar, and adjusting its level of proactiveness.

An integral aspect of Muse's functionality is its capacity to break down larger tasks into manageable plans, monitor progress, and resume operations upon detecting changes or the need for human approval. To ensure the security of user data, Muse operates on a cloud-based virtual machine called Muse Secure VM, overseen by a separate system named Sentinel.

Sensitive actions, including purchases and email correspondence, necessitate explicit user approval. Users also retain control over the services Muse can access and the permissions granted.

While Meta claims to store credentials in secure storage and prevents access to Muse data, this does not entirely eradicate the trust concerns surrounding the technology. The company has set up extensive controls around Muse, but given Meta's history of privacy controversies, consumers might view these assurances with skepticism.

The ultimate test for Muse will be whether it saves users enough time to justify granting it access to personal information and services. If successful, Meta could potentially establish a credible consumer agent; otherwise, even advanced automation may struggle to overcome the trust barrier.