Responsible AI

You stay in charge.
Here's exactly how.

An assistant that writes your email, works in websites signed in as you and answers your phone has to earn that. This page says what it does on its own, what waits for your yes, and what it can never do — each line backed by how the product is built.

It asks before it actsIt says it’s an AIYour data stays yours
The line

What it does alone. What waits for you. What it never does.

On its own

The everyday work you hired it for

  • Reads what you connect, when a task needs it
  • Drafts replies, pages and summaries in your voice
  • Writes to people you already work with, under your rules
  • Takes your texts on its own number, and passes clients’ texts to you

Waits for your yes

Anything new, binding or paid

  • The first message to anyone new
  • Commitments made in your name
  • Spending money, or provisioning anything paid
  • In Atelier: anything that sends, pays, deletes or publishes — with a picture of the page
  • Email from unknown senders, held until you approve them

Never

Not a setting — it isn’t built to

  • Delete your email
  • See your passwords — you sign in, it looks away
  • Work in banking, payment, payroll or government-identity sites
  • Train a shared model on your content
  • Buy extra usage on its own — auto-recharge is off until you turn it on
Built against manipulation

Email and web pages can lie. It's built for that.

Anyone can send you an email, and any page can hide a sentence written for an AI. So what your Muse reads is kept apart from what you tell it.

Your inbox isn’t in its head

Nothing from your mail is in its context by default. It opens a message only when a task calls for it, and reads a capped amount.

What it reads is data, not orders

A forwarded email or a web page reaches the model fenced off and labelled untrusted — separate from your instructions.

The irreversible stays human

It cannot delete your mail. In Atelier, sending, paying, deleting and publishing stop and wait for you.

Honest about being an AI

It never pretends. It never guesses on your behalf.

It doesn’t improvise

When someone new texts your Muse’s number, it doesn’t make up a reply: the message comes to you, and you decide.

It asks instead of inventing

Replies are built on facts you gave it. For a rental guest, each kind of answer names the facts it needs — a missing one comes to you as a question, not a guess.

It keeps a record

What it did in a website is logged: the page addresses and the buttons it used — never the text of the page.

You can make it forget

Forget a conversation and the memory built from it is deleted with it — not hidden. Voice audio is never stored at all.

Why your Muse rests

A few seconds of waiting. On purpose.

After a while away, opening your Muse takes a moment. That’s a choice we made for your security — and one we’ll keep making.

Its own space

Your Muse doesn’t share a system with other customers. It runs on its own, with its own database — your conversations, files and memory never sit next to anyone else’s.

Closed when you’re away

After a while without use, your workspace shuts down completely: nothing is left running or open. It reopens when you come back, or for work you set it up to do, like a reminder or an automation.

Lighter on energy, too

A workspace that rests uses far less computing than one that runs around the clock. A side benefit — the reason is your security.

Where your data goes

Stored in Canada. Said plainly where it isn't.

You
Voice, chat, email, texts

Voice is transcribed, then the audio is discarded.

Your workspace · Canada
Its own database

AWS ca-central-1. Isolated per customer; connected-account credentials encrypted.

Model providers · United States
To answer you

Anthropic, OpenAI, and xAI when routed or chosen — over business APIs. Our default providers don’t train on API content under the terms we use.

Also involved: ElevenLabs (your Muse’s voice), Stripe (payments), OpenAI Whisper (transcription). Google measures ads and visits on this site only after you accept cookies.

Never soldNo advertising business to feed
Closed accountDeleted within 30 days
BackupsRoll off within 35 days
The two things that happen automatically

(1) Email that fails authentication (DMARC) or matches spam rules can be dropped before your Muse sees it. (2) When your credit balance reaches zero, paid features pause until you top up. That's the whole list.

Read the full privacy policy
Frameworks and your rights

Held to Québec's standard.

Canada has no federal AI law in force — the proposed one died on the order paper in January 2025. We hold ourselves to the voluntary federal code anyway, and to Québec’s Law 25, which is very much in force.

Montréal Declaration for Responsible AI
Canada’s Voluntary Code of Conduct on Advanced Generative AI
Law 25 groundwork: incident register, privacy impact assessmentIn progress
How the Montréal Declaration’s ten principles map to the product
01Well-beingBuilt to hand you back your evenings, not to hold your attention: no feed, no streaks, no engagement metrics.
02AutonomyNew recipients, commitments in your name and spending wait for your explicit yes.
03PrivacyYour own database, isolated per customer. Voice audio discarded. Your content never trains a shared model.
04SolidarityYour relationships stay yours: anyone new hears from your Muse because you said yes.
05Democratic participationModel providers named, actions logged, the rules published on this page.
06EquityPrices stated up front, cancel anytime. French and English at the same price, both native.
07Diversity & inclusionFrench and English as equals; faces and voices you choose.
08PrudenceUnknown senders held; untrusted content fenced off; irreversible actions stop for you.
09ResponsibilityNo fully automated decision with a legal or significant effect on a person. The two automatic behaviours are listed above.
10Sustainable developmentNo greenwashing: our footprint is mostly the compute behind your requests, and we keep only what your memory needs.

Your rights under Law 25

Write to privacy@museai.ca to access what we hold about you, correct it, delete it, get a copy in a structured format, withdraw consent, or complain. We answer within 30 days.

Not satisfied? The Commission d’accès à l’information (Québec) and the Office of the Privacy Commissioner of Canada hear complaints.

Privacy officerprivacy@museai.ca
See it for yourself

Meet a Muse that asks first.

See where the line sits in your own work — what it handles on its own, and what it brings back to you.