Announcement: September 8, 2026 (updated September 30) · Terms updated: September 8, 2026 · Privacy policy effective: September 17, 2026 · Last verified: October 8, 2026 · We re-check this page when the documents change.
What Is Muse AI?
Muse is the assistant the internet can’t decide how to feel about — Meta’s entry in the personal-agent race, announced September 8, 2026, adopted instantly, and debated loudly ever since. This page is the plain definition: what it is, what it can reach, who’s behind it, and where its own documents draw the line.
The short answer
Muse is a personal AI agent from Meta that carries out tasks for you — inside a dedicated virtual machine, across the apps you connect, reached through the Muse app or WhatsApp. It is rolling out free across the US, it is made by Meta Platforms, Inc., and its first month generated both the strongest privacy commitments in the category and the loudest privacy controversy.
The product’s own one-line description, from the top of muse.ai:
“Try Muse, your personal AI agent that gets things done. Give Muse a goal or an everyday task and it handles the rest, from finances and health to shopping and the people you care about.” muse.ai (homepage) — accessed October 7, 2026
| What it is | A personal AI agent from Meta that carries out tasks across connected apps |
|---|---|
| Made by | Meta Platforms, Inc. — named on the app listing and in the terms |
| Front doors | The Muse app, WhatsApp, and muse.ai (Meta’s own domain for the product) |
| Access | Rolling out in the US — no invite list announced |
| Cost | Free for most tasks; subscription plans for more — terms allow free access to end at any time |
| The app | “Muse from Meta” on Google Play — 4.9 stars, 62.6K reviews, updated Oct 6, 2026 |
What it actually does
Three passages from Meta’s own documents cover the product better than most coverage does.
It acts, rather than answers. The announcement’s own framing:
“It doesn’t just answer questions, it actually does the work.” Meta announcement, September 8, 2026 — which also calls Muse “a secure, private personal AI agent that proactively helps with people’s goals and suggests ideas”
You give it goals, not instructions. The homepage’s scope, in its own words:
“Give Muse a goal or an everyday task and it handles the rest, from finances and health to shopping and the people you care about.” muse.ai (homepage) — accessed October 7, 2026
And the help center describes the leash — the sentence that separates an agent from a runaway:
“Muse is designed to ask you to confirm before it takes certain important actions like sending an email or making a purchase.” Meta help center — “How Muse handles your privacy, safety and security”
That third sentence is where the arguments live: the confirmation design is exactly what the September 2026 reporting tested, from both directions. We keep that record on our safety page, clause by clause →.
How it differs from a chat assistant
Three differences that hold up against the documents:
1 · Agent and data live in one virtual machine.
The announcement describes a dedicated VM that houses both the agent and a person’s data — Muse Secure VM — with a promised Confidential VM “later this year” that even Meta couldn’t read. Architecture as the privacy pitch is new; so is relying on it in month one.
2 · It has a front door you already own.
You can use Muse “directly in WhatsApp,” per the announcement, besides the standalone app. Every rival needs a new app adopted; Muse ships into an install base measured in billions.
3 · It’s Meta’s.
The parent company is the world’s largest advertiser, which is precisely why the documents go out of their way — “Muse doesn’t share your conversations or the data in their VM with Meta’s ad systems” — and why every privacy claim gets the loudest possible scrutiny. Both facts are structural, not incidental.
Where it stops — the September security record, the settings that matter — has its own page in this cluster.
What it can reach
What you connect is what it can reach. The help center’s word for the hooks is Connectors, and the announcement’s granularity is the part worth memorizing:
“You can choose to connect Muse to apps and services provided by Meta or third-parties. We call these Connectors.” Meta help center
“People choose which apps Muse connects to and exactly how much access it gets. … people choose what Muse can do, whether it reads their mail or can also send on their behalf.” Meta announcement, September 8, 2026
The privacy policy then sorts what it collects into four categories: your interactions (text, voice, photos), files and storage in the VM (MEMORY.md, goals, memories, scheduled tasks), information from Connectors (email and calendar activity), and activity metadata (sites visited, actions taken). How much of that surface is actually switched on is the whole safety question — which is why it has a page of its own.
Who makes it
The company is Meta Platforms, Inc. — named as the developer on the Google Play listing (“Muse from Meta”) and as the counterparty in the Muse Supplemental Terms of Service. The announcement itself carries no individual executive byline; the product is presented as a Meta release. One trust-context number from launch month: an Oppenheimer survey cited in Wall Street Journal coverage (September 22, 2026) found 8% of U.S. consumers would trust Meta with their passwords, versus about 30% for Google.
And the domain: muse.ai is Meta’s own home for the product — the announcement’s rollout sentence names it outright (“…in the US on iOS, Android, and muse.ai”), and the app’s privacy policy link points there. It is not a same-named third party.
How you get it, and what it costs
Availability, in the announcement’s own sentence:
“Muse is rolling out in the US on iOS, Android, and muse.ai, and coming soon to AI glasses.” Meta announcement, September 8, 2026 (updated September 30)
Cost, in two sentences from the same post — the split, and the escape hatch:
“It’s free for most of what people need, with subscription plans for people who want to do more.” Meta announcement, September 8, 2026
“…reserves the right, in its sole discretion, to modify or discontinue offering Muse for free … at any time and without prior notice.” Muse Supplemental Terms of Service → fees and subscriptions — updated September 8, 2026
Meta announcement (pricing) · Muse Supplemental Terms of Service → fees and subscriptions, updated September 8, 2026 · Google Play listing (in-app purchases) — checked October 8, 2026
So the accurate sentence is: free today by design, freemium by roadmap, revocable by contract.
What it isn’t
Four boundaries the documents draw themselves:
It isn’t the other Muses.
The Muse meditation headband and MuseScore the music software are unrelated products that own big pieces of the word “muse” in search. This site covers exactly one Muse: Meta’s agent.
It isn’t available worldwide.
The rollout sentence is US-only — iOS, Android, and muse.ai. AI glasses are “coming soon”; other regions have no date.
It isn’t a finished privacy story.
The Confidential VM — encrypted so “not even Meta can access it” — is announced as “later this year,” not shipped. The September security record happened on the current architecture.
It isn’t a chatbot.
A chat assistant produces text; Muse “takes action” — forms, purchases, messages — which is why its terms read like a power-of-attorney and not an app license.
Frequently asked questions
What is Muse AI in one sentence?
A personal AI agent from Meta that carries out everyday tasks for you across the apps you connect, reached through the Muse app or WhatsApp.
Who makes Muse AI — is it really Meta?
Yes. The developer on the Google Play listing is Meta Platforms, Inc., the terms are an agreement between you and Meta Platforms, Inc., and muse.ai is Meta’s own domain for the product.
How much does Muse AI cost?
“Free for most of what people need, with subscription plans for people who want to do more” — Meta’s own split. The terms reserve the right to end free access at any time without notice, and the Play listing already shows in-app purchases.
How do I get Muse AI?
In the US: the app (Google Play lists “Muse from Meta”; the download page also names Mac), WhatsApp, or muse.ai on the web. No invite list was announced — unlike rival Instinct, which is invite-based.
Does Muse AI work with WhatsApp?
Yes — the announcement says Muse works “directly in WhatsApp” alongside the standalone app. What it can do there is governed by the same Connectors and approval settings.
What’s the difference between Muse AI and ChatGPT?
ChatGPT answers; Muse acts. Muse runs in its own virtual machine with your files and memory, connects to your mail and calendar, and is designed to confirm before sending or buying. That last design — an agent with authority — is also why its safety page is longer than its definition.
Is Muse AI safe?
Layered answer, and it deserves its own page: strong written architecture (credentials Muse can’t see, an ad-system wall), a retroactive training opt-out — against a first month with a patched 0-day and a “worst for privacy” verdict in ZDNET’s testing. The clause-by-clause record is on our Muse safety page.
Sources
- muse.ai (homepage) accessed October 7, 2026 · the one-line self-description and the goals/tasks scope · muse.ai is Meta’s own domain for Muse
- “Introducing Muse: The World’s First Personal AI Agent Built for Everyone” — Meta (about.fb.com) September 8, 2026, updated September 30 · the definition, the access-granularity and rollout sentences, the pricing split, the ad-system wall, and the Confidential VM window
- Muse Supplemental Terms of Service, updated September 8, 2026 the counterparty identification and the free-revocation clause · English text verified via the Internet Archive snapshot of October 2, 2026
- Muse Privacy Policy, effective September 17, 2026 the four collection categories · English text verified via the Internet Archive snapshot of October 4, 2026
- “How Muse handles your privacy, safety and security” — Meta help center the Connectors definition and the confirm-before-acting design
- “Muse from Meta” — Google Play listing developer Meta Platforms, Inc.; 4.9 stars from 62.6K reviews; in-app purchases; updated October 6, 2026
- “Meta’s New AI Agent Is an Instant Hit—and the Backlash…” — Wall Street Journal September 22, 2026 · the Oppenheimer password-trust survey cited in coverage (8% Meta vs ~30% Google)
Every quotation on this page traces to one of these sources, with its date. Anything not quoted is our description, and says so.