Meta on September 8, 2026 introduced Muse, a personal AI agent designed to carry out digital work rather than only answer questions. The product is available in the United States through a Muse app on iOS and Android, at muse.ai, and via messaging in WhatsApp, with Meta AI glasses support said to follow.
Muse runs on Muse Secure VM, a dedicated cloud virtual machine that houses the agent and the user’s connected data. Meta says a separate Sentinel agent on the same machine gates outbound internet actions and prompts the user before sensitive steps such as sending email or making a purchase. Credentials and payment details are kept in secure storage so Muse can use them without seeing passwords or card numbers, according to Meta’s announcement.
The agent is powered by Muse Spark, which Meta calls its most capable model to date for agentic work. Meta describes Muse handling everyday tasks—sending email, booking travel, filling forms—as well as longer-running goals that continue after the app is closed, returning when status changes or approval is needed. Checkout can use Link by Stripe with one-time cards and purchase protections; Meta says Shop Pay and 1Password support are planned.
Privacy claims are central to the launch pitch: Muse conversations and VM data are not shared with Meta’s ad systems, users can opt out of training on their interactions, and a later Muse Confidential VM is planned this year so only the user holds the encryption key. Meta positions Muse as free for most everyday use, with paid plans for heavier automation.
Independent coverage from WIRED and other outlets confirmed the same-day U.S. rollout and Meta’s emphasis on Secure VM isolation as differentiation in a crowded personal-agent market. DigiEditorial treats Meta’s primary post as the authoritative product description and corroborates the launch window and privacy framing with secondary reporting.