How to Turn Off AI Training in Muse (and Other Privacy Settings to Check)

Muse can only help you if it knows things about you: your calendar, your inbox, your preferences. That's a good reason to spend five minutes deciding what it gets to keep. This guide covers the settings that matter, in order of importance.

Labels move. Meta updates Muse often. If a label below doesn't match your screen, look in Settings for the closest wording.

Step 1: Switch off AI training

By default, Meta uses your Muse conversations and the actions Muse takes to train its AI models, after removing key personal information.

To opt out, open Settings in the Muse app, find the model training setting, and switch it off.

Do this before you connect email or calendar. The setting only affects data from that point on.

Should you turn it off? That's your call. Leaving it on helps improve Muse. Turning it off is sensible if you'll connect work email, share health or money details, or use Muse for your business.

Step 2: Review what Muse remembers

Muse saves things you tell it as memories and uses them in every conversation. To review them:

• Tap your assistant icon → Identity to see Muse's memory files.
• Edit or delete anything that's wrong, out of date, or that you'd rather it didn't keep.
• Or just ask in chat: "What do you remember about me?" and "Forget [thing]."

Do this every month or so. Memories pile up quickly.

Step 3: Check your connectors

Every app you connect is another source of data. Go to Settings → Connectors and for each connected app ask:

• Do I still use this with Muse? If not, disconnect it.
• Does it need "send" or "write" access? Email can be connected as read-only or read and send. Choose read-only unless you specifically want Muse sending mail.

Disconnecting stops future access immediately.

Step 4: Set permissions to match your comfort level

Under Settings → Permissions, choose how often Muse asks before acting. Our full guide is How to set Muse permissions and approvals. The short version: keep at least "Ask for some actions", and use "Always ask" while you're new.

Also check Allowed websites in the same menu. That's every site you've told Muse it can use without asking. Revoke any that surprise you.

Step 5: Look at your activity log

Tap your assistant icon → Activity log for a timeline of everything Muse did and every permission you granted. It's the best way to spot anything unexpected.

What Meta can and can't see

Here's what Meta has said publicly, in plain terms:

Question	Meta's answer
Is my Muse separate from other users'?	Yes. Each person's Muse runs in its own isolated virtual machine.
Can Muse see my passwords or card numbers?	Meta says no. Credentials sit in secure storage, and purchases use one-time card numbers.
Are my chats used for ads?	Meta says Muse conversations and VM data aren't shared with its ad systems.
Can Meta employees access my data?	Access is restricted by policy, but Meta can access it when needed to support, secure or operate the service.
Can I download my data?	You can inspect, edit and download files in your Muse workspace, including its memory about you.
Is there a fully private mode?	Not yet. Meta has announced a "Confidential VM", with encryption only you hold the key to, planned for later in 2026.

One more risk to know about: websites can trick agents

AI agents read web pages, and some web pages contain hidden instructions meant to trick them. This is called prompt injection. Meta acknowledges it isn't fully solved. Your best protection is the approval system: if Muse suddenly asks to send something or buy something you didn't ask for, deny it and check the activity log.