Meta launches Muse, a personal AI agent that lives in WhatsApp.

Meta introduced Muse on 8 September, an AI agent that runs errands, shops with purchase protection and talks like a contact in WhatsApp; US rollout first.

Meta introduced Muse on 8 September, a personal AI agent that takes on tasks and goals for a person, powered by a new model, Muse Spark, and running on its own dedicated virtual machine, Muse Secure VM.

Talking to it works like messaging another person, in the Muse app or directly in WhatsApp. It is rolling out in the US on iOS, Android and muse.ai, is free for most of what people need with subscription plans for more, and is coming to AI glasses. It is the first Meta agent that can act on a person’s behalf across the web.

What it does, per Meta: sends emails and books travel, takes a goal and builds a plan for it, opens a browser, fills in forms and negotiates, keeps working after the app is closed and comes back for approval before sending an email or making a purchase. It checks out with Link by Stripe and is the first AI agent covered by Link’s purchase protections, using a one-time card so real card details stay hidden; Shop Pay and 1Password support are coming. It remembers details a person mentioned once, and Meta’s example is turning a recipe reel saved on Instagram into a grocery list.

The privacy claims: a separate Sentinel agent on the same machine approves anything Muse sends to the internet; Muse never sees passwords or payment methods; there is a full audit trail; people choose which apps it can reach; interactions can be opted out of model training; and Muse does not share conversations or VM data with Meta’s ad systems. A Muse Confidential VM, encrypted with a key only the user holds, comes later this year.

What it means for brands

The commerce line is the one to watch. An agent that shops with purchase protection, negotiates and remembers a saved reel is a new route from an Instagram post to a basket that a brand cannot buy an ad into, and Meta has said plainly that Muse’s data stays out of its ad systems. Nothing changes this week outside the US. What is worth doing is checking that your product pages and catalogue feeds are machine-readable and accurate, because an agent filling in forms on a customer’s behalf will read them without a human in the loop.

Sources

Written by Matt Cunnelly, edited to the NBK Social editorial standards. AI-assisted research and drafting, human-edited and fact-checked. Spot an error? Tell us.

Matt Cunnelly, Founder & CEO, NBK Social. 15+ years building social for global publishers, from UNILAD (LADbible Group) to Supercar Blondie (SB Media). Focused on the systems behind consistent, large-scale growth.

Newsletter

The NBK Social briefing

Our WhatsApp coverage, and everything else we publish, by email.

Free · Unsubscribe in one click
Subscribe