Question.Marketing

The dated AEO record

Should we be worried about Muse Charm? What Meta has confirmed about listening, training and kids

Meta hasn't said Muse Charm is always listening. The demo was tap-to-talk. What Meta has confirmed is that Muse uses your interactions to train its AI by default, and Meta calls that "a good default".

Last updated 25 September 2026

Running time 11:46Uploaded 25 September 2026

What does this video cover?

  • 01Meta hasn't said Muse Charm is always listening. In the demo, you tap a fingerprint sensor to start talking.
  • 02The Muse training setting is on when you first use it. You can switch it off under Settings > Data controls. So the deal is opt out, not opt in.
  • 03Meta calls default-on training "a good default" (Meta's Muse safety post, quoted by Business Insider, 21 Sep 2026).
  • 04Meta says Muse doesn't share your conversations with its ad systems.
  • 05Muse is for over-18s, and there is no UK launch date yet.
  • 06UK GDPR applies to companies based outside the UK when they offer services to people in the UK or monitor their behaviour here (Article 3, UK GDPR).
  • 07The UK plans to ban social media for under-16s, with the first restrictions expected in spring 2027 (GOV.UK).
  • 08The sales agent scenario is a prediction, not a Meta announcement. Simon's belief that his phone listens for ads is his personal belief.

Full transcript

The short answer: Meta hasn't said Muse Charm is always listening. The demo was tap-to-talk. What Meta has confirmed is that Muse uses your interactions to train its AI by default, and Meta calls that "a good default" (Meta's own post, quoted by Business Insider). In this episode, Simon Young of Question.Marketing explains why he thinks the default is the real privacy question, what it means for children, and how AI assistants could turn ads into sales agents that know your budget.

Read Simon's longer take in the LinkedIn newsletter: https://www.linkedin.com/pulse/always-available-talk-listening-muse-consent-who-owns-simon-young-n3nsc/. The full transcript is below.

What is Muse Charm?

Muse is Meta's personal AI agent, launched in the US on 8 September 2026. Muse Charm is a keychain device that Mark Zuckerberg showed at Meta Connect on 23 September and described as "always available to talk to". It ships in December, in time for the holidays (TechCrunch). According to Bloomberg, it has microphones, cameras and its own 5G connection. Muse is for over-18s, and there is no UK launch date yet.

Interest is high. Meta's shares rose 11.3% on 21 September and Muse went to the top of Apple's App Store (Yahoo Finance, The Motley Fool).

Is Muse Charm always listening?

Not as far as Meta has said. In the demo, you tap a fingerprint sensor to start talking. Meta hasn't said the Charm listens all the time. Simon's concern is the direction of travel, and what happens to everything we say once we do talk to these devices.

What does "training on by default" mean?

The Muse training setting is on when you first use it. You can switch it off under Settings > Data controls. So the deal is opt out, not opt in. Meta says Muse doesn't share your conversations with its ad systems.

Simon's view is that our data belongs to us, and that we have the right as individuals to stop it and opt out. In his LinkedIn post about Muse Charm (not in this episode), he puts it this way: he's not anti-AI, he's pro-consent.

Does UK law apply?

It can. UK GDPR applies to companies based outside the UK when they offer services to people in the UK or monitor their behaviour here (Article 3, UK GDPR). The ICO's Children's Code also expects high privacy settings by default on services children are likely to use. Muse and Muse Charm have no UK launch date yet, and we found no ICO comment on them as of 25 September 2026.

What about children?

Muse is 18+. Simon asks whether these devices should be allowed to interact with and listen to children or people under 18, and he thinks we 100% need controls around kids. The UK plans to ban social media for under-16s, with the first restrictions expected in spring 2027 (GOV.UK). At Connect, Meta also launched camera-free Ray-Ban Meta Audio glasses alongside its camera models (The Verge).

How could AI assistants change advertising?

Simon, who has spent years targeting ads on Google and Meta, predicts that ads inside AI assistants will become sales agents. Picture an assistant saying: "Oh, hi Simon. I see that you're trying to research golf clubs, maybe." It knows your price bracket and, eventually, has a good idea of your bank balance. That's his prediction, not a product announcement.

Why this is a knowledge race

Simon calls it a knowledge race: "a race to grab, or maintain, and hoover up as much knowledge as possible." What that means for brands is our view at Question.Marketing, not something Simon covers in this episode: if you fill a knowledge gap that isn't answered well anywhere else, you give assistants a reason to recommend you.

What we're not claiming

  • We're not saying Muse Charm is always listening. Meta hasn't said so.
  • Simon says in the episode that he believes his phone listens for ads. That is his personal belief. Instagram head Adam Mosseri said in October 2025 that Meta does not use the phone's microphone to eavesdrop on you.
  • The sales agent scenario is a prediction, not a Meta announcement.
  • Hardware details come from Bloomberg, not an official Meta spec.

Would you want your voice, and almost your thoughts, built into a training model for an AI?


Episode: "Should we be worried about Meta's Muse Charm? Privacy, consent and your voice", Simon Young, recorded 25 Sep 2026. Edited cut, 11:46. About this transcript: built from the edited video and podcast (Whisper pass on the cut audio, lightly edited for reading). It matches what is in the published cut only. Some passages from the raw recording were removed in the edit and do not appear here. One word is bleeped.


Transcript

Should we be worried about our privacy?

Simon Young: Should we be very worried about our privacy? This is something I've been talking about my whole career, because I am one of the advertising people who, over the years, has targeted people very specifically using Google Ads and Meta Ads.

I wrote a book back in 2021 with my business partner, Andy Price, talking about the privacy concerns then, and listening devices. So I was talking an awful lot about answer engines and how voice search, particularly with Alexa, was starting to impinge on, or listen to, what we are talking about.

[PAGE NOTE: the book is "Bullsh*t, Privacy, Toasters, Videos and YouTube Marketing" by Simon Young and Andy Price (FCM Publishing, 16 Sep 2021). https://openlibrary.org/books/OL62556968M]

Muse, Muse Charm and the AI assistant race

Meta is fighting with Google, with Amazon, with everybody else to become the number one AI search, or assistant. And in the US, Mark Zuckerberg has launched Muse. Massive jump in Meta's stock price off the back of it. Extremely popular in the US already as an AI assistant.

And they are then going to launch Muse Charm. That being a disc or a pendant, whatever you want to call it, that is going to sit maybe around your neck, or in your pocket, or you'll put it out on the table. The privacy concerns around these devices should really now be everybody's concern. Where are we heading with this, and who owns the data?

[PAGE NOTE: Meta and TechCrunch describe Muse Charm as a keychain device, shipping in December. See Receipts.]

Who owns our data?

What gives Zuckerberg the right to own that data? Because as individuals, that data belongs to us, surely.

Mark Zuckerberg is really, I think, taking steps to gather all of the data, right? All of the data of the entire world, or anybody that wants to have one of these devices.

[PAGE NOTE: what Meta has confirmed is that Muse AI training is on by default. Meta calls that "a good default" in its Muse safety post (quoted by Business Insider, 21 Sep 2026). Meta hasn't said Muse is always listening; the Muse Charm demo was tap-to-talk (TechCrunch, 23 Sep 2026). Muse has no UK launch date yet.]

Voice assistants and answer engines

I work with answer engine optimisation. I believe we are heading into a world where voice search will be everything, and we are moving extremely, extremely quickly. One of the people I follow, Rand Fishkin, is very vocal about the fact that search engines are changing.

And I've said we are not ending up with Google. We are going to end up with voice AI assistants. And if you look at how good things like Grok and Claude, and now Muse, are at harvesting our data through voice, it's scary. Where does this stop?

[PAGE NOTE: "harvesting our data through voice" is Simon's view.]

Should these devices listen to children?

The question I want to ask is: should these devices be allowed to interact with and listen to children, or people under 18, or under a certain age?

For children, it has now been decided that the influence of social networks upon kids' minds is so huge that we are going to have to ban it. So where do we sit with a device that listens to, and will eventually, well, talk back to?

We've got Meta glasses removing the camera, right? Because that could be an invasion of privacy. Where do we end up with these devices?

[PAGE NOTE: Muse is 18+. The UK's under-16 social media ban is expected to come into force in spring 2027 (GOV.UK). On glasses: at Connect, Meta launched a camera-free pair, Ray-Ban Meta Audio, alongside its camera models (The Verge). Meta says the timing is a coincidence.]

Living with an AI pocket device

I own a pocket device, and it's great. It's brilliant. I use it for AI note-taking. So, as an answer engine, I can pose questions during meetings. I can book meetings with people that I'm talking to. It can summarise those meetings. So it's not just recording the meeting and transcribing it. It is physically then taking the entirety of a conversation, making sense of it, putting it into a note form, putting actions in, and plans, and tasks, and talking to my AI assistants like Claude. Then that goes off to my diary, books a meeting, follow-up emails, et cetera.

So we are so quickly moving into a world where I don't know that we have a choice. My book back in 2021 was talking about your toaster listening to you, potentially, as a bit of a tongue-in-cheek thing. But I was talking about the fact that pretty much at some point in the future, every single electronic device in our world is going to be listening to us. And the saying "knowledge is power" has never, ever been more true than it is right now.

Who's learning everything?

Where does this stop? Who has got that information? Who's learning everything? I know the point of an AI assistant will eventually be to understand as much about me, or whoever the user is, as possible, and then serve them as quickly as possible what they're trying to achieve, or which answers they need, without them having to go and search on the internet. But we are seeing more information getting pulled out of social.

Influence, elections and social feeds

The UK government, Andy Burnham, our Prime Minister, was in the US this week talking about how Russia had tried to influence a previous election here, in 2019.

And I spoke to a group of young people that were working for me at the time. It was around that time. I was talking to them about what they see on social, how they were going to vote, what they see in their social media feeds. And they all eventually agreed: yes, what we see does influence how we perceive the world. Of course it does. How we vote, how our cohorts talk to each other. And therefore that influences how a generation thinks, feels and acts. And this concern over privacy should be something that everybody is feeling right now.

[PAGE NOTE: Prime Minister Andy Burnham spoke at the UN General Assembly in New York on 22 September 2026. His words on 2019 were: "we have evidence that they tried to interfere with the 2019 General Election" (GOV.UK, 23 Sep 2026; BBC News, 23 Sep 2026). See Receipts.]

When AI assistants become sales agents

These LLMs are effectively like a browser, right? They are the portal into information. And we're now at a point where we move into a privacy world, and I'm sure we will move into an area where people don't want to give their consent for their data. I would like a scenario, personally, where I'm able to stipulate: am I looking for a product? Am I looking to buy?

Because we are going to get to a point where, within the LLMs, sales agents will take over conversations. So rather than a bunch of ads, if you take Google Shopping where we stand now, you start searching for a product and you'll start seeing the product offered for sale to you in a bunch of ads. You can see the prices, you do the browsing, you don't necessarily know which of those retailers to trust. Maybe you end up on Amazon and buying a product. Why is Amazon so successful? Because it delivers it quickly. You can get exactly what you want.

That's what the LLMs are going to try and do. So they will put a sales agent in there and say, "Oh, hi Simon. I see that you're trying to research golf clubs, maybe." And it will then start talking to me about what the latest news is, what the latest products are, the trends, who's using what, what's good for my age group. It probably even knows where I go and play golf, how much I've spent in the past, what my price bracket is. Eventually, it probably knows what's in your bank account, right? It'll have a good idea what's in your bank account, whether you can afford it, whether it's worth even selling to you in the first place.

So it's pretty scary where we're going, and I don't think Mr Zuckerberg is helping.

[PAGE NOTE: the sales agent scenario is Simon's prediction, not a Meta announcement. Meta says Muse conversations are not shared with its ad systems (Meta Newsroom, 8 Sep 2026). Meta has also said that when Muse browses a website for you, it appears as your activity, so that business might use your visit to show you an ad on Instagram (quoted by Business Insider, 21 Sep 2026).]

Opt out, not opt in

I don't think we have a choice, right? The choice will be in the hands of the consumer: whether you want to be listened to, whether you want to have that information stored anywhere. And I think we have the right as individuals to stop that, and opt out, not opt in.

[PAGE NOTE: Muse AI training can be switched off in Muse under Settings, then Data controls (Meta Help Centre).]

Is the phone listening? (Simon's view)

How many years have I been listened to? My phone is on my desk while I am recording this episode of my podcast. That phone is listening to every single word that I am saying. I don't think we need to kid ourselves that it is not.

I can sit here and talk about certain products. I think we've all tested it in the past. Maybe some of you haven't. But if you sit and you talk about a product that you've never even bought or considered buying, and you talk about it with a friend in front of your phone, you will get adverts for that product.

How far that data collection goes, and how much it is assigned or attributed to an individual: I think Meta, Google, et cetera would say, "Oh no, we're not holding the data against an individual person. We are building audiences or cohorts to assign the data to, to then allow advertising to go to those people." I frankly call b******t on that, on the basis that I think they will move 100% into a very personalised system, where you and I need the option to opt out, because these things can influence your life.

[PAGE NOTE: this section is Simon's personal view and experience. In October 2025, Instagram head Adam Mosseri said: "We do not use the phone’s microphone to eavesdrop on you." (The Verge, 1 Oct 2025; TechCrunch, 1 Oct 2025). That was a general statement, not one about Muse.]

Controls for kids

And do we need controls around kids? I think we 100% do, because these devices have the ability to change the course of the world by influencing the masses.

I'd love to hear what people think about that, whether you agree or disagree.

A knowledge race

I will be talking about this a lot in the future, because clearly we're moving into a world where personalised advertising and voice search within answer engines, and answer engine optimisation, is going to become huge. And it's going to happen a lot more quickly than people can believe, because now the tech is there. It is racing. It is happening, and it is an arms race. Well, it's a knowledge race, right? A race to grab, or maintain, and hoover up as much knowledge as possible.

Your voice in a training model?

As that moves forward, it will become far, far, far more personalised, but also more intrusive.

This is the question today. Do you feel like you want your voice, and almost your thought, recorded and then harvested and built into a training model for an AI? Where do we end up? That's my question. I'll see you in the next one.


Receipts

Every factual statement on this page, in the video captions and in the page notes, with its source and date.

Opinion and prediction

These parts of the episode are Simon's own views, not verified facts: that Muse and other assistants are "harvesting our data through voice"; that his phone is listening and that talking about a product brings adverts for it; that ad platforms will move to fully personalised targeting; and the prediction that AI assistants will run sales agents that know your budget. We have not added opinions for Simon that he did not give on camera.

FAQ

Is Muse Charm always listening?

Meta hasn't said so. In the demo at Meta Connect, you tap a fingerprint sensor to start talking (TechCrunch, 23 Sep 2026).

Does Muse use my conversations to train Meta's AI?

Yes, by default. The training setting is on when you first use Muse, and you can switch it off in Muse under Settings, then Data controls (Meta Help Centre). Meta calls default-on training "a good default" (Meta's Muse safety post, quoted by Business Insider, 21 Sep 2026).

Does Meta use Muse chats for ads?

Meta says Muse doesn't share your conversations with its ad systems (Meta Newsroom, 8 Sep 2026). Meta has also said that when Muse browses a website for you, it shows up as your activity, and that business might use your visit to show you an ad on Instagram (quoted by Business Insider, 21 Sep 2026).

Can children use Muse?

No. Muse is for people aged 18 and over (Muse terms and the Apple App Store listing, via Daily Caller News Foundation, 10 Sep 2026). Separately, the UK's social media ban for under-16s is expected to come into force in spring 2027 (GOV.UK, 15 Jun 2026).

Is Muse available in the UK?

Not yet. Muse launched in the US on 8 September 2026 (Meta Newsroom), and as of 25 September 2026 we found no UK launch date or UK price for Muse or Muse Charm.


Would you want your voice built into a training model for an AI? Tell us what you think.

This entry is part of a dated archive, what we said, when we said it. For what works in AI search today, start with AEO at Question.Marketing. For why the dates matter, see the receipts.

Where next?

The library holds every video we have published. The AEO pages explain the method behind them, what it costs and what the evidence actually supports.