Meta's New AI Agent 'Muse' Sparks Global Excitement and Privacy Concerns

Meta has launched Muse, an autonomous AI agent operating in the cloud that can perform tasks like making purchases and sending emails directly from WhatsApp. The system, which runs on a secure virtual machine, has generated significant buzz in the tech world, quickly topping the App Store charts in the US with approximately 600,000 downloads in its initial days. This viral success has reportedly boosted Meta's stock by over 21% since its release.

Muse operates differently from AI agents requiring dedicated hardware. It runs within a secure cloud infrastructure called Muse Secure VM, powered by the Muse Spark model. Users can interact with Muse through a dedicated app, on Macs, or directly within WhatsApp. The AI can handle complex tasks such as filling out forms, booking travel, or converting an Instagram recipe video into a shopping list. Integration with Stripe's Link wallet allows Muse to make online purchases using single-use virtual cards. A key feature is its ability to continue operating in the background even after the app is closed, breaking down large tasks, negotiating, and contacting service providers for inquiries, only returning to the user for final approval on sensitive actions like payments or sending emails.

Despite global enthusiasm, Muse is currently available only in the United States, with Meta actively blocking access for users in other countries, including Israel, even when using VPNs. However, some Israeli users with U.S. accounts have reported successful operation, suggesting a global infrastructure is ready for expansion. The launch has also raised significant privacy and data security concerns due to the AI's access to emails, messages, and computer files. Meta emphasizes its security measures, including the secure virtual machine and a separate control agent that blocks unauthorized internet access. They state that conversation data and access details are not shared with Facebook's advertising systems, and an end-to-end encrypted version will be released later this year.

A recent incident where a tech journalist suspected Muse scanned private Mac messages without explicit permission caused a stir. Meta engineers clarified it was a misunderstanding of the bot's response wording, not a data leak. This event highlights public apprehension towards tech giants and their AI capabilities. Muse's immediate popularity indicates the arrival of the personal AI agent era, but its long-term success hinges on Meta's ability to assure users about the protection of their privacy.