Orchestrate all the Things - Fair Patterns: Designing for Human Autonomy in the Age of AI. Featuring Marie Potel-Saville, Fair Patterns Co-Founder & CEO

Episode Date: June 1, 2026

We live in a world fraught with interfaces designed to work against you. Most people sense it. Few can prove it, and fewer still are trying to fix it.  There is a particular kind of frustration that arrives when you understand a problem perfectly and cannot do a single useful thing about it. Marie Potel-Saville spent years living inside that frustration. First as a competition lawyer cycling through antitrust litigation across Europe. Then as in-house counsel watching the law get treated as a cost center. She pivoted, trained in innovation by design, and launched her first company in 2018. Potel-Saville was looking for the gap between what the law said and what the internet did to people. She found it immediately. The term was “dark patterns.” And her first instinct – the instinct of a lawyer and a builder – was to ask a question nobody in the field had thought to ask: what’s the antidote? That question eventually became Fair Patterns, a multimodal AI platform that detects online manipulation and addictive design at scale. Fair Patterns went live in January 2026, and recently won the 2026 Digital StartUp award in cybersecurity and sovereignty. It is trying to redefine digital products and online experiences. Article published on Orchestrate all the Things: https://linkeddataorchestration.com/2026/06/01/dark-pattern-detection-how-fair-patterns-uses-ai-to-fight-back/

Welcome to orchestrate all the things. I'm George Anadiotis and we'll be connecting the dots together. Stories about technology, data, AI and media and how they flow into each other, shape in our attacks. So very simply, the dark pattern is an interface that makes you do something you wouldn't have done if someone had asked you the question. If you had understood what's going on and if someone had asked you, you would never have done that action. It's a very simple definition and it's quite ironic because in our R&D lab, we studied hundreds of scientific articles, very, very sophisticated on all the various definitions of dark patterns. There are 16 taxonomies and ontology. I mean, the level of sophistication of scientific research on dark patterns is really impressive.

But we've decided to go back to a very simple definition because that's what manipulation is about. You are manipulated when you're doing something that you wouldn't have done otherwise, right? And that's why we believe that there's a strong need for another type of structural remedy, which is human safety tech per se. It's a bit of an irony because we need tech to protect ourselves from the tech. But it's, you know, that's the world we live in. We live in a world fraud with interface is designed to work against you. Most people sense it.

Few can prove it and fewer still are trying to fix it. There's a particular kind of frustration that arrives when you understand the problem perfectly and cannot do a single useful thing about it. Marie Potel Soville spent years living inside that frustration. First, as a competition lawyer, cycling through antitrust litigation across Europe. Then as an in-house council watching the law get treated as a cost center. She pivoted, trained in innovation by design and launched her first company. company in 2018. Hotel Seville was looking for the gap between what the law said and what the internet did to people. She found it immediately. The term was dark patterns. And her first

instinct, the instinct of a lawyer and the builder, was to ask a question nobody in the field had thought to ask. What's the antidote? That question eventually became fair patterns, a multimodal AI platform that detects online manipulation and addictive design at scale. My name is Marie. I'm the co-founder of Fair Patterns, a multi-model AI that scans everything that's live to detect online manipulation and addictive design. And what brought me there is a journey that started as a private practice lawyer. I was in competition law for 10 years, in big law, in London, Brussels, Paris, you know, during a multi-emerger filings and antitrust litigation. And actually, interestingly enough, the competition law lens

is something that I use every single day in my work currently, which makes me very happy, by the way. So anyway, I did that for 10 years. Then I felt that my brain was a bit shrinking because of the super technical aspect of what I was doing. It was really, really niche. And I felt like, I don't know, only a handful of people really understood what we're doing. And so I felt my brain was shrinking and I felt the need to go to the other side to go in-house because I wanted to broaden my activities. So first, I lived in Mexico City for three years as an expatriate, had two kids, worked there in a law firm.

And then back to Europe, I decided to move in-house. And this is where actually I could broaden my activities that was achieved. However, I also got to measure the huge gap between the law and what happens in reality, which you tend not to see in private practice, but it really hits me in-house. And it really made me sad because I believe that the law is so important to create. create value. It's really a value creation center. And in-house it tends to be seen as a cost center, as, you know, the business fund police, just a constraint and nothing more. It made me really sad. And so I trained. I did a master's degree in innovation by design. And I took a leap of faith, creating my first company back in 2018. I'm reassuring myself that, you know, if it didn't work, I could always go back and find a normal job again, which is absolutely not true. But that's how I reassured myself.

So anyway, I became an entrepreneur. My first company was called Amorabi, and it had a mission of bridging the gap between the law and its users. And interestingly, in the middle of this journey, I came across the notion of the notion of, of addictive design and dark patterns. And immediately, my first reaction was, what's the antidote? What's the opposite of addictive design? What's the opposite of a dark pattern?

And no one had an answer. And so I started to fix it, but it was very manual. It was very human-led on a project-by-project basis. Obviously, I realized that the scale of the problem implied a technology solution to match that scale. And so I first created the R&D lab for Fair Patterns, where the research question was, how can we fight at scale against dark patterns?

And from there, we built Fair Patterns as an AI solution. So it was first an R&D lab, then we created the concept of a fair pattern, which we published. And it means interfaces that empower you to make your own free and informed choices. It sounds obvious, but it's actually the opposite of what's happening right now online.

We're just Muppets. And so from the concept, we created the library of fair patterns, and then we created the multi-model AI that scans, sites, apps, and social media. And we've been live on the market since January of this. Yeah. Okay. Thank you. That's quite a fascinating journey, I would say. And thank you for anchoring the different parts of that journey. And we'll have the chance to dive into lots of

those throughout the conversation. And I think for me, an obvious place to start would be precisely going back to what you mentioned, but to the thing that caught your attention about addictive design and the so-called dark patterns. And we'll come to the antidote part later. But first, I think, you know, in order to understand the antidote, we have to first understand what the disease is, so to speak. So what are dark patterns, actually? To be honest with you, so personally, I was vaguely familiar with the term,

and it sort of re-emerged when I saw the work that you are doing. I was like, ah, aha, yes, there is this thing called, dark patterns, but it was a long time ago. And I think for many people, it's the same. They may actually experience it every day, but I'm not sure they have a clear definition of what it is and why it's a bad thing, actually. Yeah.

So very simply, the dark pattern is an interface that makes you do something you wouldn't have done if someone had asked you the question, if you had understood what's going on. And if someone had asked you you would never have done that action. It's a very simple definition and it's quite ironic because in our R&D lab, we studied hundreds of scientific articles, very, very sophisticated on all the various definitions of dark patterns.

There are 16 taxonomies and ontology. I mean, the level of sophistication of scientific research on dark patterns is really impressive. But we've decided to go back to a very simple definition. because that's what manipulation is about. You are manipulated when you're doing something that you wouldn't have done otherwise, right? And so, yeah, this is the way we reframe it right now. And as you said, you know, most people experienced it probably even several times a day, from cookie banners to, you know, free trials or free delivery that end up as a very much paid for a subscription,

subscription that you can't cancel to delocation that you know you are very much incited to to authorize under a false pretext and you end up being totally tracked so that's a dark pattern it's really manipulation and deception through the interface through the architecture of choice that's online think of it as choice engineering so that it's no longer your choices it's no longer your decisions you think that you do stuff online as you want you actually don't because there's this architecture of choice engineering your behavior so we're just Muppets basically and that's different from addictive design of course okay so I want to

share like a personal let's say experience just to make because in my mind There's a kind of thin line, let's say, separating guided guidance that a designer may offer through the user experience, through manipulation. So my personal experience is this. So I like to use, for example, web-based, let's say, email providers, not necessarily through, I don't know, Outlook or a dedicated mobile application. I like to use them through the browser. And so every time when I open, for example, Gmail through my mobile browser, every time I get the welcome screen, it has a huge button, upgrade and use Gmail application.

I specifically don't want to do that. Luckily, there is an option. It's very tiny hidden underneath. You can still click here to use it through the... So I do this every time. So obviously, you know, the... people who designed this user experience want you to use the web application, the mobile application. They still give you the option to use the mobile browser, but it's

tiny, hidden and you have to specifically look for it and you have to go through that every time. So they're clearly steering you towards this direction, but you still have the option. So this obviously is their preferred way of interacting with the application because they want you to download the mobile application for a number of reasons. So my question is this, would you classify that as a dark pattern? And if yes, is that actually in different jurisdictions? Is that legal or is that a kind of gray area? Yes, so this is totally a dark pattern because it's making something more difficult

because it's not in the interest of the company for you to not use the app. Okay, so your preferences are not respected. And instead, they're trying to steer you towards the solution that fits their interests and not your interest. And the reason behind is usually that they collect way more personal data when you're using the mobile app, obviously. And this is where it enables me to answer your legal question. Because if, you know, the data collection process is forced, then it's not valid consent. You know, if they trick you into sharing more personal data, then you would have done if you had really understood,

then it's contrary to GDPR, because consent has to be free and fully informed. And also, you know, it could be contrary to GDPR in terms of, there's a principle that's very often completely forgotten and overlooked. You know, it's the data minimization principle. It's completely forgotten because we live in the attention economy, obviously. But what GDPR imposes to companies is to collect the minimum personal information that really need to do their job. Not the maximum, the minimum.

And so, yeah, there are many grounds under which it's actually illegal. Okay. So it's interesting that you brought up GDPR. because again to my mind it's kind of the archetypal example of something that was maybe well intentioned but has ended up being well sidesteps in a lot of ways so for example the the principle of collecting as little personal data as possible for example is in practice counteracted by something else which is called legitimate consent so people and companies end up doing lots of things that were not normally under legitimate consent. Well, actually, it's either consent or legitimate interest.

It's one or the other. And yeah, well, for legitimate interest, let's not forget that companies have to do an impact assessment. Okay. And the interest is only legitimate if it's absolutely indispensable for them to collect. those data right so you can't you know you can't just bypass consent if your data protection impact assessment doesn't enable you to use the legitimate interest you know it's not like an open door to collecting your anything you want how you want it right so I presume that there is no specific legislation for for dark patterns it just falls under legislation that exists independently such as GDPR for

example let me clarify so you've got two types of legislations in europe and then we'll leave the the US aside so the first type of legislation is principled based it doesn't specifically prohibit dark patterns but it applies to dark patterns so that's the case for GDPR for example most dark patterns will be contrary to the fair-nast principle that's enshrined into the GDPR where any processing of personal data needs to be fair, you know, and if you're tricked into giving your data, not knowing what's happened, it's just unfair. So that's how GDPR applies and also obviously, you know, the valid consent conditions. Then you also have UCPD, the Unfair Commercial Practices Directive.

you know, it's a general prohibition of aggressive commercial practices and misleading commercial practices and dark patterns fall into that category. Okay. So it's, it applies, but it doesn't specifically prohibit dark patterns unless you take the appendix. But anyway, and then you've got a second type of legislation, which is more recent. And that's the DSA Digital Services Act, the DMA, Digital Markets Act, and the DMA Digital Markets Act, and the Digital Fairness Act, which is still at the stage of a project, the DSA prohibits expressly dark patterns in Article 25. This is an express prohibition.

The DMA has an anti-circumvention provision, specifically on dark patterns, and the AI Act specifically prohibits dark patterns in AI in Article 5. And then we've got the draw, well, the project of a digital, Fairness Act which will fill the gaps for dark patterns and prohibit addictive design more precisely because it's imperfectly covered currently in EU legislation. Does that make sense? Okay, it does. It does. I just wonder, does it actually happen in practice that people

take companies to court on these accounts? Absolutely. I'm sure you've heard about the meta YouTube verdict on March 25th. So in this case, a jury held meta and YouTube liable for addictive design. So for precisely designing their platforms, their social media in a way that makes teenagers and kids addicted to their platform. By the way, the World Health Organization recognized a social media disorder. So it's not an addiction like you can be addicted to a substance, to a drug. It's the same damaging mechanism. So it's been officially recognized as a social media disorder.

And yes, so you've got, you know, beyond the meta YouTube verdict, which is a landmark case. It's the first time in history that platforms have been held liable for a digital. predictive design. And what's really interesting is two factors. One, in this case, the discovery process in the US shed the light on internal documents within Meta, which they had to produce before court, which are so telling. Basically, they were in a competition race with TikTok. They were losing the race. And to avoid losing the competition race, they found out that the younger the kids started using social media, the more addictive it would get.

And so you've got one internal document saying exactly that, that if they want the product to be as sticky as possible, they need to target kids at an even younger age than 13 years old, which was the official age limit. And so you've got a quote in one of their internal documents. If we want to win big in teens, we need to take them as twins. And twins means, you know, pre-adolescent. It's actually kids from eight to 11 years old.

And so there was a conscious, very conscious, very well-informed decision with a quote from a document from Zuckerberg himself saying, okay, this is our strategy to win the competitive race against TikTok. we need to make young kids a target. And so this verdict is also interesting for another reason because you've got thousands of very similar cases pending. So it's only the first of a very long series. But now that there was a guilty verdict, it's very likely that all the thousands of pending cases

will have a similar... They can use it as precedent. Yeah. So that's in the US. Now, in the EU, we're also seeing a number of class actions against companies employing dark patterns. For example, there's one currently against booking.com in the Netherlands. That's ongoing as we speak. And we know that there will be additional class actions against addictive design in the very near future.

Okay. Okay. So it does happen. from the user or consumer, however you want to put it side of things, that's encouraging. But I want to sort of dive a little bit deeper here in the sense. You also mentioned in the introduction that people have been very thorough in studying and classifying the different types of dark patterns. And to my surprise, it was only when preparing for our conversation that I discovered that they even went as far as to produce taxonomies of dark buttons. And actually I discovered that you yourself have worked on one of those.

So this is of special interest to me and I'm presuming to many of the people who may also listen to the conversation because I have an interest and background in data modeling and semantic data models, taxonomies, ontologies and so on. So I was wondering if you could share your personal experience, let's say, as well as the experience of others in the field that you have worked with. So how they went about creating those models and what has their utility actually been? Have these taxonomies been useful for you? Yes. So, okay, where to start? First, when I started researching dark patterns and I found out that there were already

16 taxonomies existing. So some of them were created by researchers, scientists, some others were created by regulators themselves. You've got a taxonomy by the FTC. You've got another one by the EDPB, specifically on privacy. I was really impressed because the, yeah, the sophistication of the research was, yeah, was really impressive. actually, after three years of R&D, you know, we spent three years analyzing all of this, including the ontology that was created by Colin Gray, Christiana Santos, and Natalia Bielova, who are amazing researchers in this field. I have a huge admiration for them. But what we realized after having analyzed all of this is that ultimately and quite ironically,

the sophistication of the taxonomies and even the ontology doesn't help to cure the problem. You know, it's intellectually very interesting. It makes for great scientific papers. But you're still left with no practical solution and with a sort of digital plague that keeps spreading as you develop, you know, the conceptual framework to understand what it is. I mean, it's a necessary work, but you're always running behind the reality, which is that as we speak, new dark patterns are being created. Right? And that was even before Gen AI. And now with Gen AI, you've got AI powered dark patterns.

Okay, so Gen AI is actually creating dark patterns, whether you're prompted to do it or not. You can prompt it to do it and it will do it super efficiently, but you also have AI dark patterns. within LLMs themselves, you know, which takes the problem to a whole different level. So my experience with taxonomies, and again, I'm really grateful for the ontology, but it left me with a sense of frustration, also because I'm a doer. You know, I need practical solutions. I need things that really change the reality for people. And this is where I thought, okay, we can.

contribute to this to this field by actually creating a countermeasure because it's it was a bit surprising to me to see the state of the art with so many papers on okay what's the name of this problem what's the root cause of this problem what are the legal grounds that we could use to fight against this problem and no one ever looked at a solution i even had this discussion with harry brignall who kindly agreed to be our senior advisor you know he's the the expert who coined the term dark pattern and then have it, he had it evolved into deceptive patterns, deceptive design. And I even had this discussion with him. I was like, okay, you know, where's the antidote? And at the time he honestly told me that he didn't think it could be cured. He thought,

you know, he sort of assumed that this was a sort of side effect of digital economy. Like, you know, this is a tension economy and there's no other way. And this is something that I strongly rejected, probably because I'm a lawyer and because I believe in the rule of law and I can't accept that something so deeply illegal on so many legal grounds is a sort of side effect and we can't do anything about it. Also, because as a citizen, I was very concerned about, you know, this manipulation system, you know, it's bad enough when it makes you buy a pair of sneakers or that you didn't need or pay, you know, your plane seat. Or get a subscription you didn't intend to or whatever of these things. It's bad enough, you know, in the commercial area. But think about the political

implications. Right? We already had a precedent with Cambridge Analytica. I mean, there's scientific evidence that Brexit perhaps might not have happened without that manipulation on social media. And it can happen again, and it's probably happening again as we speak. So, you know, this is also why I just couldn't accept this idea. And so my experience with all this very interesting intellectual work is also that it's not enough. And it's nice to be intellectually satisfied, but you really need a cure. You really need a solution, a practical solution,

live on the market. Absolutely agree, and we'll come to the solution part. Just a short comment on your experience in actually using and contributing to this ontological and taxonomy work. There's one famous Adas in the world of data modelers and an ontologist. It's that, well, I think it perfectly aligns with your experience. It basically goes something like, well, you can create like, you know, a perfect ontology and, you know, a very deep taxonomy and all of these things.

But you should always start with the end goal in mind, like, okay, what is the use case? What am I supposed to do with that? Who am I supposed to serve? What are the questions that I need to answer? all you're doing is documenting, then it's a kind of intellectual exercise with no real utility. And I think that has been your experience, probably. I fully agree. To be fair, the ontology was created by Colin Greene, Christiana Santos and Natalia Bielovai,

with a very specific purpose in mind. Basically, they had seen in various cases in the US that some companies employing dark patterns would take advantage of the fact that taxonomies were fragmented, you know, it was a bit scattered all over the place. They used this to create an argument that is confusing, we're not sure what a dark pattern is, to get away with their behavior. And so they decided to create an ontology

that would encapsulate all of the taxonomies precisely to counter that, you know, sort of legal argument that it's a vague concept, we're not sure, and if it's vague, you can't prohibit it. So this one had a very specific purpose. But yes, I fully subscribed to the adage. What are you trying to solve? Who are you trying to serve?

Exactly. So, all right. So let's say that one way of fighting back is the legal way. So you explained what kind of legislation applies to dark patterns. You explained also that there now is precedent. So in theory, if I am, if I get to the point, that I'm super annoyed, for example, with my provider that makes me click that obscure link to use the mobile browser every time. I could take them to court. But I think that most people

will not go through that process simply because they don't have the time or the funds or the energy or anything. So companies get away with that most of the time, unless they cross this threshold that they do things that are extremely, I would call, I would say evil, like getting consciously addicting, for example, twins in the examples that you mentioned, classifies beyond competition or market practices. So these things should definitely be prosecuted. But for the lesser, let's say, evils, is there a way to fight back? And I think this is where the fair pattern concept comes into play. And I would like to explain what constitutes a fair pattern and what you are trying to achieve with that. Yeah, well, I completely agree.

with you. Most companies, you know, get away with it because regulators have limited resources. You know, they prosecute the most visible cases, but it's never going to clean up entirely the market. And then class actions obviously exist in certain jurisdictions, not all of them. But it's also down to the resources that a few NGOs and perhaps some VCs that fund the litigation cases could dedicate. So it's really important in our mind to empower people to fight back and to provide what we tend to call a human safety layer, you know, that should be embedded in everything that's digital and everything that's in all AI products. So what's a fair pattern? First, in, you know, conceptually, it's an interface that empowers you to make your own free and informed choices. If we break that down, it means that, you know, contrary to dark patterns that exploit your cognitive weaknesses to make you buy something, pay more, subscribe when you didn't mean to share your personal data, when you didn't want to. Or contrary to addictive design, which creates, you know, dopamine addictive loops where, you know, you're really no longer in control of your use.

of the digital product, the social media, it's just doom scrolling and you're not even enjoying it. You know, there is scientific research showing that it's not because the teenagers spend all their time on social media that they feel good about it afterwards. They feel very bad about it. You know, they've got headaches. They feel bad about themselves. And in the case of the meta YouTube verdict, this young girl, She even had, you know, suicidal thoughts, dysmorphia, terrible symptoms. So contrary to what I now call predatory design. I've made up this term recently after the verdict. I now, you know, encapsulate dark patterns and addictive design under predatory design

because it's really what it does, you know, preying on the cognitive weaknesses of the people it is supposed to serve. So a fair pattern does the exact opposite. It takes into account of cognitive weaknesses. For example, we are all very sensitive to information overload. So, you know, if I put you in front of a screen absolutely full of content, like, I don't know, a privacy policy in legacies, you know, in full screen, like a world of jargon, I can be 100% sure that you will never read it because no human would read it.

It automatically triggers a number of biases because information overload is one of the triggers where plenty of biases come into play and make us stop reading, stop learning, but automatically clicking nevertheless. So to be fair, a pattern has to acknowledge that humans, all humans, are actually vulnerable online. We are very vulnerable to manipulation because of all these cognitive biases. For example, we have the default bias. It's absolutely obvious, but you know, when there's, when something is pre-selected, like a box that is pre-ticked, very few humans are going to un-ticket, very, very. We just leave it as is. So it's just a very basic example, but in a case in the UK, it led to unwanted purchases in the amount of billions of pounds. And so a fair pattern

takes into account the fact that all humans are vulnerable online. In some cases, for example, for minors, it also specifically takes into account that, specific vulnerability. For example, before the age of 25 years old, the prefrontal cortex is not entirely mature. It will do so only at 25. And it means that minors, teenagers and kids, don't have the means to fully control their impulses. So if you engineer the stimuli that will make them addicted, they have no way to fight against it. That's why social media and video games when there is addictive design. It's so, so, so sadly efficient on them.

So back to fair patterns. It takes into account human vulnerabilities and it leverages behavioral sciences and design, design strategy and plain language, legal plain language, to provide just the sufficient amount of information, but not too much. because then you're overloaded and you don't read. So enough information so that you can make an informed choice,

but not too much information so that you're not overloaded, taking into account your needs. And, you know, in terms of design, it's all about providing the actual means of the choice. You are referring earlier, you know, to the number of clicks that you have to do to not download the app when you don't want. want to. Well, typically, a fair pattern has symmetrical buttons to do the action or reject the action. You know, yes, no, accept or reject all. Symmetry in terms of shape, text, so that, you know, the text doesn't try to dissuade you from doing the action that the company doesn't

want you to do, like reject the promotion or the offer. So that's for the design. It's really about symmetrical options and providing the actual means of a choice. It can be as simple as a button. You know, sometimes you just don't have the button to reject cookies. You have a big, fat button to accept all and a tiny gray link in a corner of the screen that you don't see with a very low contrast. So all of that needs to be fixed by design. And then the legal plain language comes into play because we're also very overwhelmed by legacies. There are very interesting studies showing that when people are faced with legacies, so legal jargon in a very authoritative style, it also shuts down their learning process and reading process.

And so the more I so plain language, the better, we're lucky to have a plain language ISO norm. norm since 2023 and a legal one, so legal plain language since last summer in 25. And that also really helps people to

actually understand the tiny portion of information that we provide. And the whole point, by the way, of fair patterns, is not to nudge users. We don't believe in nudging. Because nudging is

internalistic, you know, who decides what's good for you or for me? So we don't want to nudge anyone, but we want people to fully understand what's going on and then make their own decision. If they're happy to accept all cookies because they use it, you know, they like to have super targeted ads because it serves their purpose. Let's say it's easier to shop online. If that's what they, if they really understand how it works. And if If that's what they want, then fine. But we want people to fully understand and then make their decision. Now, for the case of addictive design, a fair pattern is a bit different.

It's still based on the same principle to take into account the cognitive weaknesses of humans. So typically our vulnerability to dopamine shots. And then a fair pattern is more about removing those addictive features and providing creating designs where you can decide, for example, upfront, what's the amount of time you want to spend on social media? So, you know, let's say when you open your social media, you know, if it's fair by design, you could set up a time. you say, okay, I want to spend 15 minutes. You set up your time and then you've got a pop up or something when your time is up. That's just one example where you're more in control and you make a conscious decision. I like social interactions.

I like to see what's new today. I make that conscious decision and I enjoy the digital product. But I know where to stop and I'm the master of my own decisions. Does that make sense? It does. I just wonder what you're describing basically goes against the interests of the companies that are currently implementing. The ones that are implementing dark patterns are doing it for a reason. For the reasons you outlined earlier very vividly through that internal memo, for example.

So I wonder, okay, so this is great in theory, but how are the reasons? they ever going to adopt something like this? So let me take my competition lawyer hat on. I'm no longer a member of the bar, but I practice competition law for 10 years. And this is my competition law analysis of what's going on. Basically, you know, normally competition law is supposed to be, well, market economy is supposed to be the best model that we've found so far for the optimal allocation of resources.

Okay? That's the theory. That's the dominant economic theory. There are a number of conditions for that to be true. And one angle is that market economy provides this optimal allocation of resources and the best outcomes for consumers. When you break it down, best outcomes for consumers, it means better products. So more quality, more innovation, lower prices, better services.

Okay? That's the dominant concept under which we accept market economy. I'm not the one saying this. This is in all the OECD papers and like 100 years of economic theory, basically. Now, what happens, especially with addictive design, even more than dark patterns, is that it completely undermines all of the... all of the references that we have for market economy. Because if you take, for example, the fact that people stay a lot on social media,

if you looked at it, you know, from a competition low angle, it looks like people are enjoying the products. So it's an optimal allocation of resources. So it's perfect. Except it's a total corruption of the system because people are not actually choosing to stay on social media. They're addicted to. that product, right? So it completely blurs all of the references that we have for analyzing what's a healthy competition and competitive market. The other thing that I really want to say is that,

you know, if a market stops serving end users, consumers, but actually starts preying on them, exploiting them systematically, you know, this is not an accident. This is systemic exploitation of humans' cognitive weaknesses. Then it's not serving its purpose anymore, even from a sheer economic standpoint. And I add that in 2022, the OECD published a report on commercial dark patterns, and they warned that there will be a very substantial lessening of competitiveness, if companies ended up competing not through the quality of their product or better innovation or cheaper prices, but only through the efficiency of their dark patterns.

And so for addictive design, you know, it's systemic exploitation of cognitive weaknesses. For dark patterns, it simply prevents people from switching. In computational, switching is absolutely key. When, you know, for example, when prices go up in one product, people can switch to another product that's cheaper or they decide to stay because they find the more expensive product better in terms of quality or something else. Dark patterns prevent the ability to switch. If you're, you know, trapped into a subscription that you can't cancel, it completely erases your ability to switch. Right? So that's directly anti-competitive. So it completely distorts competition at the root.

And so when a market does that, well, it's no longer a properly functioning market. And there has to be a systemic answer to that. Okay. So if I get the, the gist of the argument you are making correctly, you're basically saying that it's that some regulatory authorities should intervene, should the conditions that you're outlining be met? So basically you're saying that this is anti-competitional behavior and therefore the regulator should intervene. Am I correct in interpreting this way?

Yes, but not only. I do believe that this is a very serious anti-competitive behavior that requires competition authority to move forward. So far, we've had consumer protection authorities with sanctions that are not very effective. GDPR, it's a bit more powerful, but competition authorities have been a bit scarce in terms of enforcement.

And I really think that it's very much needed and justified. But let's be honest, regulators only, even competition authorities, they also have limited resources. What's interesting, by the way, with the competition authorities, is that they can impose structural remedies. You know, it's not just about the fines.

The fines can go up to 10% of the worldwide turnover, so it's the highest ceiling that we have among, you know, the AI Act, DSA, GDPR, etc. But fines will never solve this alone. What's interesting is that they can impose structural remedies. So this is the next frontier for competition law. But that won't be enough either. And that's why we believe that there's a strong,

need for another type of structural remedy, which is human safety tech per se. It's a bit of an irony because we need tech to protect ourselves from the tech. But it's, you know, that's the world we live in. And so we believe there's, you know, a bit like we've taken the habit of ensuring cyber security. So for example, you do penetration tests before launching any product in terms of you know, your cybersecurity framework. Well, in the same way, we believe there's a need for a human safety assessment before any digital or AI product goes live to make sure that it doesn't contain manipulation

or addictive design, because that's predatory and that really, you know, backfires to the people it's supposed to serve. Okay, so this gives me the perfect segue to. ask you what I've been meaning to ask you for a while actually. So how does all the theory, let's say, of the fair patterns apply to what you do in your day-to-day job in the organization that you have started, which is actually also called fair pattern. So based on the few bits of information I've seen myself, it seems that a big part of what you do is actually you have developed, this is what you also just refer to your own technology.

to identify dark patterns. And this is fine, but how does that actually translate to take in action? Yeah, sure. So basically what we do is that we identify dark patterns and addictive design on anything that's live, so sites, apps, social media, video games, etc. And we can do the same in mock-ups,

so before it is life. But we don't just identify it. We provide a fair pattern. And so when we work for companies, when they are our client, basically they come to us because they're not sure how to identify everything that needs fixing on their digital properties. They want to be extensive. They don't know how to do that. And they're not sure how to fix it. You know, you know what's wrong. You're not sure what's right. So that's what we do for companies. For example, HP is using our Figma plugin so that their designers can see. scan their mockups to ensure the absence of dark patterns before any new page goes live on their website. So that's ongoing. Then we also do the same for influencer marketing. That's another use case that's really interesting. We haven't talked about it yet, but there are plenty of dark

patterns and addictive design on platforms like TikTok shop, YouTube, and it's really pushed by influencers and sometimes very ironically brands are paying influencers a fortune and the same influencers are actually employing those completely illegal techniques and it's actually the responsibility of the brand as well you know it's not just the influencer and so we have a use case where we have brands to police the videos that are produced by the influencers that they pay because you know they pay them and then the influences do their stuff and that's it. So we scan those videos and we automate the messages to the influencers in agreement with our client obviously so that it can be, you know, just suggestions of addition or notice and

take down if it's really too serious. We automate the reports to the clients and we can also automate the reports to the influencer agencies so that they also learn and better manage their influencers. So that's what we do for companies and it helps cleaning the market either at the markup stage or when it's already live. And again, we've only started in January, so it's early stage, but we've got like six clients for now. And the second use case is law firms.

with two situations. One, they want a digital audit for their clients. And so we just provide the technology so that it's way more efficient. And it provides the remediation for their clients. And second is the strategic litigation where we provide packaged evidence for any litigation law firm that wants to go after one of the companies employing dark patterns. And obviously, we have a very strict conflict. of interest policy so that we can't be on both sides.

If I may, I want to add that we're currently working on integrating our solution in a very famous Gen A.I company, I can't name it for now. But ideally, you know, we want to fix the problem at the root. So integrating within Gen A.I. itself would be absolutely amazing so that any line of code produced by, you know, this tool would have this human safety layer embedded. Okay. We also want to do the same in other technological platforms, like we're talking to a number of potential partners for now.

We could integrate in, you know, Cloudflare, Replit, lovable. We could integrate in consent management platforms, you know, so just for cookie banners, but that would be really nice. again, again, you know, add the root to fix the problem even before it produces damages. Okay, so I wonder on the entrepreneurial, let's say, side of things. Obviously, you started really, really recently, but I imagine that even before you publicly launched, you must have gone through the usual, let's say, cycle of, well, going from an idea to building a team and fundraising and all.

of this so it sounds to me like what you're doing probably falls under the not just creating a product if you want to call it that but also a category is that really the case or are there others in the field doing something similar to what you are doing thanks a lot for noticing this we're actually creating a new category of human safety tech we're the first mover in this category but we really we see this category moving there's is not a competitor, but there's a very interesting company called DeValience that just launched a product called Spectre 1, which cancels the noise so that any device around you cannot listen to your conversations. You know the reality that you end up being spied upon by your

microwave or your connected home, you know, the series and the like. Well, they created a device that makes your conversation completely shielded from any device that can be listening to you. And so I think, you know, there's a movement to go in that direction. You know, human safety tech is clearly catching up because the, you know, I think the Gaffams just went too far in exploiting us. and we need to fight back and now is the time. It does, it does. And I wonder how hard or easy has it been for you to go to prospects and convince them that this is something that they should be doing. So, of course, being an entrepreneur is incredibly hard and it's absolutely not credible to say anything otherwise.

But, you know, I've got two co-founders, amazing. Jerry Gupta, co-founder and CTO, and Sunder and Rianan, co-founder and chief AI ethics officer. And I really want you to do a tribute to them because we're more than co-founders. We're a sort of extended family. And yes, it's incredibly hard. But what really makes me happy is that the market is responding really well. So it's tons of efforts. It's super difficult. It's, you know, crazy hours and everything. But the market is responding really well. When you have a client as big as HP in the US, and we're talking about the headquarters, saying, wow, this is exactly the tool we've been dreaming about. And we even, you know, you didn't even know it existed.

you know, this clearly it signals that something is happening. And also what makes me happy and hopeful is that we have many people who contact us, not necessarily because they're in a position to be our client, but just because it's usually very senior people with amazing positions in very large organizations. And they're like, what can I do to help your project? I swear that I get at least, I don't know, a handful of these, of these, you know, contact requests on LinkedIn every single week. Like, what can I do to help? I'm, you know, for example, someone from MIT, someone who held a super senior position at the White House previously, not the current administration. You know, people like that who really understand how evil the tech has become.

So it's usually, you know, very knowledgeable people and they're just offering to help. And so I think that we're creating way more than a product and even way more than a category. There's a movement being created. There's a community of people around the world. That's also very interesting. You know, people from Australia, Japan, Latin America. It's really, really interesting. that are deciding to fight back

because again, the degree of exploitation has really gone too far and they're ready to fight back. So what can actually people do to help if they do want to help? I mean, besides the obvious spread in the world. And I would also take the question one step further. So can you also provide some kind of help to people? let's take a hypothetical scenario that I am an employee at Company X and I'm not happy with the way that they do their design. And I would like someone like you to do an audit on them and go to my boss and tell them, hey, look, you know, you should stop doing that.

Is that a possibility? Of course. Yes, of course. And we're always happy, you know, to have discussions, very open discussions. we know that change is hard and we don't judge anyone. Sometimes we even found that dark patterns are not even intentional. It happens plenty of times in our audits. And so, you know, do come to us. We will have an open discussion. It's a solution that's also very affordable.

It's easy to implement. And we will also help managing the change. That's our promise. And yeah, if anyone is unhappy, you know, they can always reach out to us. I do my very best to answer to all of the messages on LinkedIn. It's a full-time job, but I do it. You know, do reach out either on the, through the site, there's hello at FP or on LinkedIn. We do our very best to answer to all of the requests.

And we have a very practical, easy-to-implement solution. And I might add that we are having economic data showing that it's actually more profitable to bet on transparency and trust rather than tricking people with dark patterns or addictive design. So it's also, you know, it makes a lot of sense commercially too. Okay, well, I would love to see that data by the way and just add them as reference to the article if nothing else. else. Sure, I'll send you the article. Well, it's a series of articles actually. Yeah, we've got about 10 economic papers showing that. Okay, great. I actually wanted to ask you a little bit on the on about the technical side of things as well. So how did you develop? What I presume is specifically trained and fine-tuned model of your own to do it to identify the dark patterns. But I think it's

I don't know if you have the time for that. Maybe just a couple of minutes because I've got a lot of work for the evening. But yeah, so we didn't develop our own models. We've got seven patents pending on the way we orchestrate our AI agents. So, you know, we take available models depending on, and, you know, there are new releases every single week almost. So we take the best ones depending on what we need to identify. So it's not the same model for the text and for images and for videos, etc. But basically, I think our specificity is the orchestration of the hundreds of AI agents that we've created depending on context.

So we've created a methodology that's being patented to first identify the context and then we've. call upon the most relevant agents to detect very at a very granular level the type of addictive design or dark patterns. That makes sense in this particular context. And that makes for an accuracy rate of 85%, which is above the state of the art, which is roughly 50, 60%. So maybe that all that work that went into the taxonomies was not a total waste after all, because It sounds like it's used in some way to identify the specific type of dark pattern every time. Yes, you're right. Ultimately, what really creates the change on the market is not the name of the dark pattern.

You know, is it a roach motel? Is it a subscription trap? That doesn't change anything. What does change is the legal mapping. And this is something we've done. You know, connecting a specific. type of pattern with like 10 different legal violations in US and EU law. That's what convinces the clients, the companies to actually change the design.

Exactly. Great. Thank you. Thank you very much for a super interesting conversation. And I hope everything goes your way because I think pretty much everyone stands to benefits. Thank you so much. Thanks a lot. Thanks for sticking around.

For more stories like this, check the link in bio and follow link data orchestration.