The Edit is our weekly show, where Social Editor Grace Robinson quizzes editor-in-chief Ben Hanson on some of the most significant fashion and technology stories from the past seven days.
This week on The Edit, Grace Robinson and Ben Hanson discuss Getty Images’ planned bankruptcy and the future of editorial photography, AI accountability as brand equity, ticket bots and Meta Muse’s personal dossiers, electronic circuits spun into washable fibres, and the physical consequences of retail cyberattacks.
This show is available in video format, embedded above, as well as audio, on Apple Podcasts and Spotify.
For deeper interviews with brand and technology leaders, our prestige interview show releases on Thursdays everywhere you find podcasts – and will be coming to YouTube soon.
Subscribers to our mailing list will receive notifications for both shows.
Grace Robinson: Welcome to The Edit from The Interline. The show where we run a quickfire analysis on our pick of the most important fashion and beauty technology stories from the last seven days.
I’m Grace, the social editor, and I’m joined by Ben, the editor-in-chief. And together, we have around twenty-five minutes to give you our analysis on the stories that we think really matter.
Ben Hanson: Good morning, Grace. Good to see you. I’m dealing with an uncharacteristic problem in Manchester where there’s too much sunlight. So I’m trying to filter it out in a way that doesn’t shine directly on my face, but also doesn’t sacrifice the fact that it’s a really nice day and I want to bring some of it into the room.
You said it was raining in London, so I am going to enjoy the fact that for once, the north of England has better weather than the south.
Grace Robinson: Yeah. It’s a good problem to have, so I think you should make the most of it.
Ben Hanson: I’ll take it while it’s here. We’ve got a busy week. We’ve got a bunch of AI, some non-AI stuff. Let’s get rolling.
What Getty’s crisis means for trust in images
Grace Robinson: A huge story that we have to talk about this week is Getty’s bankruptcy filing. So they were meant to merge with Shutterstock, which is one of their main competitors, but because of Getty’s $2.1 billion worth of debt, this merger didn’t go ahead. So Getty is now going to be private, and Reuters did a really great piece on it this week and they were saying that it was because of this debt, but it was also because of the rise of AI imagery that Getty has had this downfall.
Now, we do know that Getty tried to sue Stable Diffusion a few years ago, and I do personally think that it does seem very likely that AI is the reason for this downfall. So I wanted to get your take on this story and I also wanted to know what you think this means for the future of kind of AI-generated editorial imagery for fashion or fashion publications in general.
Ben Hanson: Yeah. It’s a sad one and a complicated one. So Getty, we’re talking about Getty Images specifically. So one of the primary kind of photography agencies that the biggest news organisations in the world turn to, along with a stock photography library that’s been available to corporates for a long, long time. One of the primary ones, as you said, alongside Shutterstock.
Also, one of the main holdouts, I would say, to the idea that a photo should represent, like, a ground truth. When you see a piece of photography attached to a piece of editorial, the implicit promise of it was that a person was physically there with a camera made up of a lens and a sensor. They pointed those things at a real thing that was happening, and what you licensed as an enterprise or a media outlet, and what you saw as a reader or a consumer was a representation of reality.
I don’t think the business model behind Getty has been super stable for a while. Like, media is going through a lot, and I think that definitely affected it. So I don’t think it’s quite as simplistic as AI is the death of photography, but I also don’t think it’s a million miles away from that.
Shutterstock still exists. Shutterstock forged a six-year agreement, I believe it was, with OpenAI in 2023, to actually license their library of stock photography for training purposes for OpenAI’s image models. At the time, that was DALL-E. It’s now GPT Image, as it’s called. GPT Image 2.5 is the current one, whereas Getty took a slightly different approach.
So you mentioned that they tried to sue the company behind Stable Diffusion, Stability AI. That was not fully rejected. I think there’s still some appeal process happening in England, in the jurisdiction of England and Wales, but broadly speaking, that case did not succeed on its merits.
I’m testing my memory slightly here, but I believe Getty then did sign some kind of agreement with OpenAI, but it was a separate one. It was a display agreement rather than a training agreement. So that means that they permitted their images to be shown in search results and in generated outputs to the best of my knowledge. So if you search for a news story within ChatGPT, you get a roll-up of news stories from across the web, and you get some images put your way. That’s the display agreement rather than the training agreement.
Now I think a lot of people are going to point to this as, like, AI’s Kodak moment, if you want to call it like that. Everybody loves that reference frame where Kodak, the company that made a lot of physical film, a lot of celluloid, failed to identify the scale of the move to digital photography, and didn’t capitalise on it, and then went out of business. It’s the standard business case study for not recognising the perils of a new technology as an incumbent.
That played out over, like, fifteen years or something along those lines. Think from the advent of digital photography to Kodak eventually saying, oh, dang. We’ve really misplayed this.
The timescale here is very, very different. So you’re talking about a shortfall from Getty’s public listing in 2022 to today, where they’ve lost about 90% of their public value and have been delisted from stock exchanges. That’s four years. It’s incredibly quick. I don’t think you can put all of that down to business model perils. I don’t think you can put all of that down to a mismatch between supply and demand or anything else. AI is absolutely a factor in this.
If you are a media outlet and you want images that do not necessarily have to be ground truth for the news, you’re not covering the news, you just want editorial content, it’s cheaper, it’s faster, it’s more controllable to do things with AI than it is to license them from stock libraries.
Getty didn’t, I think, have the same problem that other stock libraries have, which is they slowly themselves became infiltrated by AI-generated content. We saw that happen with Adobe Stock Library. We’ve seen it happen with other places.
I think this has a lot of bearing for what happens in fashion now. Not necessarily because fashion companies were licensing images from Getty, more that the market, the consumer, has been accustomed to that idea that if you see a picture, it represents reality, because that’s what authority is.
We had the combination of, like, Getty Images, the Associated Press, Reuters, these kinds of, like, big institutions. You looked at their words and their pictures, and you’re like, that’s the thing that really happened in the world.
This news suggests that we are past that. It suggests that people are now going to be inundated with stuff that is a representation of what happened, but is not a photograph of what happened. That changes, I think, quite quickly how people feel about images in general, that they are shown by brands, they are shown by retailers, they are shown by other organisations.
I don’t want to sound too dramatic, but this to me feels like the tipping point where we are in, like, a post-photography, post-image world. And I’m sure fashion brands, which we’ll talk about in a minute, will push back against that. Fashion consumers will push back against that. But the overall cultural zeitgeist for me only goes one direction from here, which is AI imagery, AI photography becomes the norm, and that has a big impact on a lot of the pushback and a lot of the concerns and things that we’ve seen in fashion at the moment.
How long will AI accountability shape brand trust?
Grace Robinson: Yes. So on the topic of AI, I did see a really interesting piece in Vogue Business this week. And it was written by a data scientist called Rumman Chowdhury. And she was explaining how now AI accountability can actually equal brand equity for fashion brands.
So essentially, as you were mentioning, there is so much scepticism and fear around AI and how brands are using it. So now in order to retain consumer trust, brands need to be really clear about where they’re using AI, why they’re using it and also where they draw the line in terms of where AI isn’t necessary and where they want to still use human-led creativity or human processes.
And the piece was saying that again, in order to keep consumer trust, brands really need to think strategically and think about where their audiences and consumers will accept AI and where they won’t accept it. So I wanted to get your take on this and what you think fashion brands need to be thinking about with AI accountability.
Ben Hanson: I’m going to trot out my fatalist cynic’s garb again for this one based particularly on what we’ve just talked about with Getty Images. Sure. I think this is transient, though.
I think this is something you need to do as a brand. You need to have an AI policy. You need to say, okay, this is where we will allow it to touch our creative output. This is where we’ll allow it to touch our products. This is where we’ll allow it to touch our retail experiences, our customer engagement and support, and so on.
I don’t see the necessity for that lasting more than a couple of years. I think AI adoption is swift enough, and AI rollout and kind of dominance of sectors like imaging, like we’ve just talked about, is going to happen quickly enough that this will be one of those cases where people don’t stop caring, but resign themselves to something that has been foisted upon them very, very quickly.
I think it’s the same thing that happened with social media, but on a more accelerated timescale. Yes, everybody is up in arms about how much social media has hijacked their attention, or monopolised their data, or what have you. There was a little bit of backlash to oversharing and stuff in the early days, and then people just very quickly got on with it.
Now, I don’t think that’s a good thing. I don’t think it’s a good thing in this case either that companies will just have effective kind of carte blanche from the market to go, yeah, just use AI wherever you want. I don’t care anymore.
I think that sort of resistance is something the brands need to be aware of between now and, let’s say, 2028, 2029, like that sort of time frame. Clip this, play it back because I could be horrendously wrong. But
I think by 2030, labelling requirements and stuff still exist. But this idea that AI accountability aligns with brand equity, I think, ages quickly. But that’s the simplest way that I can put it.
And I did read this article, and it clearly comes from a place of expertise, it comes from a place of passion, and so on. It talks about kind of brands bringing their internal process champions and subject matter experts and things into the AI process, like helping them to understand what it can do, what it can’t do, what they want to press it into service to do, and so on.
Cool. But again, brands are commercial enterprises, and I think in a fairly short period, the steady, like, maturation of frontier models means that AI is doing more of that work over time, and I think the people involved in helping to train it, the people involved in helping it to fit into their companies will maybe righteously say, oh, did I just train my replacement for a bunch of this? I just don’t think it’s this simple, and I certainly don’t think it’s something that lasts for very long.
What happens when everyone has an AI agent?
Grace Robinson: The next story is actually pulling together two stories that you wanted to talk about this week. So the first one is that the music artist Raye has openly pushed back against ticket touts using AI. So to put it into perspective, she had a concert that only had 5,000 seats. But because of all of these ticket touts using AI technology, she had nearly two million sign-ups to this for the pre-sale of the tickets.
So as a result and as a pushback to this, she’s now resorted to manually handpicking fans and then manually sending them sign-up links so that they actually get the chance to get the tickets.
The other story that you wanted to pair with this was about Meta’s new agent Muse. So there was a great report in Time that was showing and shining a light on the fact that Muse now keeps a very personal dossier on each user. It includes personal information, but it also includes personal information about the user’s relationships.
So in theory, your information could be in someone else’s Muse even if you don’t use Muse. So these are two very interesting stories, but I wanted to know why you wanted to put these together and what’s interesting about them.
Ben Hanson: Yeah. I think rightfully when I flagged these up, you’re like, ‘What the hell is he on? Why are we going to be talking about these things on a fashion technology podcast?’ I will get into that.
I will say, as somebody who hasn’t been to a lot of concerts myself, but has tried to book concerts for my oldest, man, tickets are expensive anyway, and tickets are incredibly difficult to get anyway. That’s not for a fashion technology podcast.
This one is because in both of these cases, we’re talking about agentic AI acting on somebody else’s behalf. Correct? So in the first case, it’s acting on behalf of, I’ll call them bad actors, like people who want to acquire tickets at face value and then scalp them afterwards and sell them on for a higher profit.
That’s existed for years and years and years. There are so many lawsuits that have gone Ticketmaster’s way for these kinds of monopolistic practices and sort of criticism that’s gone artists’ way for not allowing regular people to basically access tickets anymore.
Now that’s all automated. Right? It’s been automated for a little while in the sense that if you were a scalper, you could create a bot army, and you could go do this kind of stuff. It took a bit of setup. It was fairly clandestine.
When you equip regular people with always-on agents that, in the case of Muse in particular, are free, nobody’s paying for the compute for these kinds of things, you kick this into a significantly higher gear. Right?
If I wanted to sign up for tickets for something and there was a pre-sale, and I had an agent working on my behalf, there’s nothing stopping me saying, ping this portal every second until you get through and check out as quickly as you possibly can. None of that is anywhere outside the realm of possibility for the moment.
Now you could apply that to a bunch of fashion models. The first one is anything drop-related, anything hype-related. I know that kind of model has fallen out of favour a little bit, but anything that requires people to wait in line to sign up for pre-sale to any of that kind of hype-driven model.
I’m hesitant to call this the end of that, but it feels like very, very close to the end of that. The same way we’ve talked about AI fast-tracking the end of photography in some sense. I think we’re looking at agentic AI fast-tracking the end of these kind of scarcity-driven, get people through the door kind of business models, because that technology is just available to anyone to use. And it doesn’t just have to apply to ticketing. You can apply it to anything that requires you to sell a product.
Then if you look at the Muse story that you mentioned, so obviously, your personal agent is going to build up a dossier on you as part of the promise. Right? The whole deal with OpenAI and Meta Muse is that they’re able to take action on your behalf. Siri AI. It’s not very good, but it’s the same concept, like what they refer to as personal context. The more it knows about you, the better able it is to be an advocate for you in whatever sense that means.
People made a big kind of furore around the idea that Muse was going to cancel a bunch of subscriptions on their behalf, which I think did end up happening, take part in class action lawsuits for them in the US, just the kinds of stuff that you can do if you have access to somebody’s email, someone’s personality, and a bunch of information about them.
We talked last week about ASOS and the amount of kind of personal search history and things that were potentially being leaked as a result of the cybersecurity incident there. What Meta’s doing with Muse is that, but legal, or at the very least within the terms of service.
So Facebook was famous back in the day for building up what they referred to as like shadow profiles, so you didn’t have to have a Facebook account. You just had to be known by a sufficient number of people in your social orbit for Facebook to have what they refer to as a shadow profile of you, which covered your interests, your age, a bunch of other things about you, all of your social connections and so on.
And when you did eventually acquiesce and sign up for a Facebook account, hey, look, all that social graph, all that knowledge graph, and everything is already populated for you.
Do that with shopping preferences. Do that with search history. Do that with relationships with people. And all of a sudden, when you sign up for Muse or when you sign up for something else, you don’t have to have been explicitly parting with any of this.
So in the case of ASOS, that search history was because people were signed in on the website, and they were doing searches. It was permissioned to some extent, and it was voluntary to some extent.
In this case, you don’t even have to be using Muse. You just have to know somebody who’s using Muse for it to start capturing preferences and to start capturing personal information.
And when you slide into it, if you eventually go, okay. Fine. I will use Muse. It probably creates a better experience in the sense that it knows more about you off the bat.
All of this agentic swirl, they’re putting these capabilities in people’s hands, allowing them to know more about them than ever. Huge impact on fashion for the business model stuff, the drop thing, but returns.
So many returns don’t get made because people can’t be bothered. Soon as you start to have an AI agent that advocates for you, and you reduce all of that friction, you’re like, buy this thing for me, buy it in five sizes, and manage all the returns for me as well.
And if the returns are not free because the company feels like I’ve abused the returns policy, get on with their support agent who is also AI, and keep hammering them until you get a satisfactory resolution. I am not paying for these returns. That is how people will feel about this.
And putting that power in people’s hands and allowing that power to grow around their direct hands to where it cascades around to the people either side of them: just a huge shift in the internet and a huge shift in how fashion lives on the internet.
I don’t know. This one, to me, I feel like there’s something really big happening to the internet that does scare me a little bit. How do you feel about it?
Grace Robinson: Feels like every day AI capabilities ramp up, and it makes you just want to turn the internet off altogether, to be honest.
Ben Hanson: Don’t do that because, yeah, I would very much like people to keep reading and listening to The Interline, but I understand the sentiment.
Could our clothes become the circuit?
Grace Robinson: So the next story we’re going to talk about is a new development in electronic yarn technology. So, essentially, electronic yarn technology is where you embed a small digital electronic circuit into a yarn, then this is made into a fibre and then this can be made into a garment.
This week we’ve seen a new research paper from Nature Electronics and they found a new way to embed electronics into a yarn that’s more washable and more flexible.
So to me, this seems like it has a load of really good applications or could potentially have a load of really good applications for sportswear. So a garment could track your temperature or your muscle activity or even pollution around you.
It could also have really good implications for healthcare. So with vital sign monitoring and stuff like this. So I wanted to get your take on this, and do you think this is a really interesting development?
Ben Hanson: I do. There’s a really quick answer to this one. You’ve always been able to put sensors in fabrics. So it’s always been the case that you could say, okay, I will integrate some electronics in the sense of temperature sensors, moisture sensors, that kind of thing particularly into compression garments and so on.
There was that famous Levi’s and Google Jacquard project where they embedded something slightly more complex in a rigid denim jacket that allowed people to, like, swipe on the sleeve, I think it was, to activate something.
This is different in that what they’re proposing is embedded circuitry within a single yarn rather than just a sensor or rather than having to have other electronics in there. So it’s being able to have a whole integrated circuit end to end in a single yarn fibre.
Now that’s interesting because what that opens up is a lot of what they refer to as, like, signal processing or logic operations, like being able to do some measure of compute somewhere between, let’s say, your hand and your shoulder, or some measure of signal processing somewhere along your body.
I think it’s going to take time to see how this plays out. This kind of stuff doesn’t get commercialised very quickly. But we talk about wearables a bunch on the show. We talked about the Meta Muse charm. We’ve talked about smart glasses. We’ve talked about whatever’s coming from OpenAI and so on as fashion categories. I think that’s right. I think fashion is right to look at individual deployables like that as a fashion category.
I think the longer-term view is actual circuitry and actual computation to some extent being worked into the garments that we wear.
If you think about, like, VR headsets and things where they have to have multiple cameras facing downwards and outwards so they can see you pinching or they can see you doing that kind of stuff, like, that hand control stuff is magical, but it’s still very crude because it just relies on that. It relies on actually observing the external part of you.
If you’re able to sense that kind of stuff and you’re able to send those sorts of signals, like, directly along the garment, then all of a sudden that opens up a ton of possibilities for, you’re running, like, gait analysis. You’re reaching for your smart lock. Like, there’s a whole bunch that you can do without having to have a dedicated device in there.
So my kind of story, interesting to think about, going to take a while to commercialise, but I will be keeping an eye on it.
Retail’s cyber risks extend beyond customer data
Grace Robinson: So we’re ending this week with a really interesting story. So I know last week we spoke about ASOS’s cyberattack but this week there’s been a cyberattack that’s impacted a physical store.
So earlier in the week, almost 700 supermarkets in France experienced a cyberattack on their refrigerator systems and this caused their physical refrigerators to heat up and this obviously spoiled some of the food and also even some of the refrigerators actually set on fire. So it could have potentially been very dangerous.
Now, a lot of people off the back of this have said that it could have been prevented by very simple cybersecurity measures. So introducing a VPN or even just changing the login credentials of the refrigerator system online.
So I know this is about supermarkets but it does show that cyber hacks could have a very detrimental impact on technology that’s in physical stores. So what do you think that fashion needs to think about this story?
Ben Hanson: Yeah. And again, I can give you a very quick answer on this one. First is, as somebody who really likes cheese, the idea of people purposefully spoiling vast amounts of cheese is crazy, but kind of interesting in that regard.
To me, this is a testament to just how much connected technology there really is in retail. So if you look at some of the news coverage of this, you will see that the fridges and stuff have the programmable logic modules and controllers that govern all the temperature and the lighting and things like that.
There are also the kinds of ones that have the little e-ink screens for pricing and things like that as well, which presumably are remotely and centrally updated rather than somebody having to manually go around and set a price with little buttons every time they put a different product on stock.
When you think about the camera networks that exist in retail, when you think about the RFID tags and similar sorts of things that monitor the movement of products around stores, when you think about warehousing and picking and logistics and distribution, there is so much physical infrastructure for retail and for fashion that I am willing to bet most companies don’t know how secure it is.
Now, if you work with big retail tech providers like an Avery Dennison or an Impinj or something like that, I’m sure they have impeccable security credentials. But as we saw from the ASOS hack, a lot of this tends to happen when you have big networks of third parties all providing different pieces of the same puzzle.
And for me, for fashion, what that means is brands and retailers needing to kind of apply real scrutiny to their risk surface, not just as it pertains to the kind of personal data that we talked about last week with ASOS, but as it pertains to their actual physical infrastructure across their store network, across distribution and logistics and so on.
Now there’s nothing to suggest that this was an AI-enabled hack, but at the same time, this seems like exactly the kind of thing that somebody would say, I want to cause some havoc. What’s something that might be insecure? Help me probe this. Help me figure this kind of thing out.
We’re going to see a ton more of this. And it’s going to continue to probably be in grocery or somewhere else. But we know that fashion retail, M&S, Harrods, ASOS, these kinds of companies have been targeted specifically because they have a lot of consumer information, but they also have a lot of physical presence, particularly in the case of, like, multinational or kind of multi-city, you know, big brands, big retailers that are present across an entire country, across a bigger geography.
Their risk surface is not just data. Their risk surface is their physical infrastructure as well. And just because the story is about cheese doesn’t mean it’s something that fashion should ignore because it doesn’t take a big leap for it to become about clothing or beauty.
