Meta's Muse Agent Handles Email, Travel, and Shopping with Sentinel Approval
Meta's Muse enters the AI agent space with a safety architecture that isolates each agent in its own VM and gates real-world actions through a permission agent, addressing the core risk of AI making purchases and sending messages.
Reporting from 1 source: GIGAZINE.
Meta announced Muse, a personal AI agent that sends emails, books travel, and makes purchases on the user's behalf. It runs on the Muse Spark model and operates in a dedicated virtual machine per user, with a separate Sentinel agent that checks actions and requests permission before any transaction. Muse is now available in the US on iOS, Android, web, and WhatsApp. Japan release timing is undecided.
During the announcement, Meta demonstrated Muse planning a December trip to Japan, building an itinerary for Tokyo, Hakone, and Kyoto, booking a restaurant, and then finding a 5 km running course near the hotel to add to the user's training plan. The agent also handled a stroller purchase, showing product comparisons and a final Allow confirmation before payment. Payments run through Stripe's Link service, with disposable card numbers so the actual card number is never passed to the seller. Shop Pay and 1Password support are planned.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.