Meta opens Muse personal agent in the U.S. on iOS, Android

Meta opened Muse, its consumer personal AI agent, in the United States today on iOS, Android, and muse.ai, with AI glasses support promised later. The product is built on Muse Spark — the same closed model family behind Muse Spark 1.3 — and is pitched as an agent that does work across a person’s apps, not a chat box that only answers.

What it does: Muse runs on Muse Secure VM, a dedicated cloud virtual machine Meta says holds the agent plus that person’s data and credentials. People message it in the Muse app or in WhatsApp. Meta says it can send email, book travel, fill forms, keep working after the app closes, and ask for approval before sensitive steps such as sending a message or making a purchase. Checkout can use Link by Stripe (one-time cards; Meta calls Muse the first agent covered by Link purchase protections). Meta says Muse does not share conversations or VM data with its ad systems, and users can opt out of training use and disconnect apps.

What it is not: This is not an open-weights drop and not a global GA. Rollout is US-first. Meta’s Confidential VM (user-held encryption keys) is listed for later this year, not today. Safety and isolation claims are Meta’s.

Who it is for: US consumers Meta can reach through Instagram/WhatsApp habits who want an always-on helper with app connectors — and who will trust Meta with email, calendar, and payment rails after the company’s recent privacy and youth-safety fights.

Risk: A personal agent with browser, payment, and messaging reach concentrates compromise and prompt-injection risk on one VM. Meta’s Sentinel gate and per-action approvals are the stated mitigations; treat them as architecture claims until outsiders test them.

What’s next: EU/international dates; whether free-tier token caps throttle real agent use; App Store / Play review friction; any early abuse or payment-fraud reports; Confidential VM ship date.

— Rowan Keel