#36 - Your Face Belongs to Lensa

In November of 2022, an app called Lensa exploded across social media, celebrities, influencers and millions of ordinary people, uploaded their selfies and got back stunning AI-generated portraits. For $7.99, you could see yourself as a space warrior, a renaissance painting or a cyberpunk hero. It was magical, it was viral, and at its peak, Lensa was making almost $8 million per day. But behind the magic was something darker. Women were getting hyper-sexualized without consent. Artists discovered their work had been stolen to train the AI. Users didn't realize they had just given away the rights to their faces, forever. This is the story of how a Russian copycat app wrote the AI wave, engineered a perfect viral loop, extracted $70 million in 60 days and disappeared before anybody could stop them. It is a story about timing, exploitation and what happens when AI needs unregulated capitalism. It is the dark side of AI, not beautiful and strategic, like AlphaGo that you spoke about a couple of episodes before. But this is morally messy and ruthlessly profitable. Now, I had to add a disclaimer that this represents my analysis and commentary based on the publicly available information and reporting. Alright, so let's get back to our story. This is December of 2022, and your Instagram feed is certainly full of them. What is them you ask? Your friends, your colleagues, celebrities, all posting the same thing, AI generated portraits of themselves, stunning, artistic and impossible to ignore. This is your co-worker as a warrior princess, your college friend as a cyberpunk hero, a celebrity you follow rendered in the style of a Renaissance painting. The captions all say the same thing, lensei AI with a fire emoji. You think, that's cool. I want to try it. You download the app, you upload 10 to 20 selfies, you pay $7.99. And a few minutes later, the app gives you back 50 portraits of yourself in styles you have never imagined. It feels like magic. What you don't know, what almost nobody realized in that moment is that you just participated in one of the most ruthless, wealth extraction schemes in tech history. You gave away the rights to your face, women unknowingly consented to having their image sexualized by an algorithm. Artists had their lives were stolen without compensation and a small team in Russia just made $8 million in one day. By the time anybody realized what was happening, lensei had already extracted over $70 million and vanished into the regulatory fog. This is the story of lensei AI, the story of how speed, timing and moral flexibility created a fortune and why it matters for everything we are building in AI today. Welcome back to the identity navigator. This is your host, Rohit. Let's go back to the beginning. To understand how lensei one, you need to understand who they beat. In October of 2022, just one month before lensei exploded to indie developers launched an app called Avatar AI. Their pitch was simple. Upload yourself e and we will use AI to generate stylized portraits of you. Fantasy warriors, sci-fi heroes, artistic renderings, all powered by technology, constable, diffusion and this is basically an open source AI model that had just been released to the public a few months earlier. You and I could have used it if we wanted to and Avatar AI was good. It worked and people liked it. They made a few hundred thousand dollars in their first month. Now imagine this. We are not talking about big corporations. These were two indie developers and for these two developers that is a huge win. Like a few hundred thousand dollars in first month. That's crazy money. So they were proud and they thought that they have found product market fit. Around the same time around another app launched profile picture AI. Same concept. Upload selfies, get AI portraits, also doing decent numbers. So those were the first move words, the pioneers. They saw the opportunity early, built a product and got to market first. But here is the thing about being first. You don't always win. Because while Avatar AI and profile picture AI were celebrating their moderate success, a team in Russia was watching, studying reverse engineering. And they saw something. The first move was, didn't see. They saw the viral look. Now let me tell you a little bit about Prisma Labs. They are a Russian company founded by Alexei Moisinkov and is more team of engineers. They have been around since 2016 and their fame or claim to fame was an app called Prisma. A photo filter app that you might have heard of, that used neural networks to turn photos into art. It was moderately successful. Millions of downloads. But by 2022, it had faded. Now folks, all of us know the app store is brutal. Yesterday's viral app becomes today's forget and relic. So in late 2022, Risma Lab was looking for their next hit. And then they saw Avatar AI. They saw profile picture AI and they realized, this is it. This is the wave. But they just didn't copy the product. They optimized it for virality. Here is what they did differently. First, mobile first design. Avatar AI and profile picture AI were fine, functional. But Lensa was beautiful. The UI was smooth. The onboarding was frictionless. Every screen was designed to get you from download to payment as fast as possible. They understood that in viral moments, you have about 30 seconds to convert someone before they lose interest. So they optimized every pixel for speed. Brilliant if you ask me. The second thing that they did differently was immediate paywall. Avatar AI let you try the product before paying. Lensa didn't. You uploaded your selfie and then bang paywall. 799 for 50 avatars limited time offers. This is a dark pattern. You have already invested time uploading your photos. You are emotionally invested. You want to see the result. They are asking for a crazy amount of money. And so you pay. And here is the genius. They didn't just charge 799 dollars. They buried it in a subscription model. The default option was actually a 3599 per year subscription disguised as a free try. Most users didn't realize they had just signed up for a recurring charge. The third thing that they did differently was viral mechanics was built into the product itself. When you got your avatars back, the app made it effortless to share them. One tap and your portrait was on Instagram, Twitter, X, TikTok. And every shared portrait became an ad for Lensa. Think about the loop. I share my portrait. My friends see it. They want their own. They download Lensa. They share their portraits. They are friends see it. The loop compounds by very definition of viral post. This is textbook viral engineering. And Prisma Labs had done it before with their original Prisma apps. They knew how virality work. And the fourth thing that they did was the timing. And this is the most important part. Lensa launched their magic avatars feature in late November 2022. Do you know or can you guess what else launched in the late November of 2022? Chad GPT. Chad GPT launched on November 38, 2022. It exploded. Suddenly everyone was talking about AI. AI was everywhere. The entire cultural side-case shifted to what What can AI do? And lens I was sitting there with a prop. that answered that question in the most shareable ego gratifying way possible AI can make you look amazing so they didn't create the AI wave they wrote it so in November of 2022 Lensa launches magic avatars at first it's slow a few thousand downloads some early adopters trying it out but Prisma lap is invited they know how well it works they just need this park and then on November 30th this park happens MKBHT or Mark is Brown Hill one of the biggest tech influences on the planet post his lens avatars millions of followers seat and the floodgates open within 24 hours Lensa downloads spike 631 from 219,000 downloads on October to 1.6 million downloads in the first week of December December 5th Lensa is the is the number one app in the app store do you know who were the other apps that Lensa beat it was ahead of the Q2 ahead of Instagram ahead of TikTok celebrity started posing their avatars influencer post theirs your co-worker post theirs your mom post hers everyone wants in and here is what happens behind the scene Prisma lap is printing money at 799 per user with 1.6 million downloads and an estimated conversion rate of 22 30% they are making millions per day some estimates put their peak revenue at 8 million dollars in a single day not bad for a day's work so let me say that again 8 million dollars in one day and for context most successful apps make that in the year Lensa made it in 24 hours over the next 60 days Lensa generated an estimated 70 to 75 million dollars in revenue this is one of the fastest wealth extraction in tech history faster than floppy bird faster than pokemons go early days faster than almost anything before it and it is all happening in a regulatory vacuum no one is checking if this is ethical no one is asking about consent no one is questioning the terms of services the app is just taking taking your photos taking your money taking your data and scaling exponentially so now everyone is posting their Lensa avatars it's fun it's viral it makes you look beautiful cool and it feels harmless but then people started noticing things the first problem was sexualization many uploaded normal selfies professional headshots casual photos family pictures started getting back avatars that were inappropriate hyper sexualized revealing clothing that they that wasn't in the original photos poses that felt suggestive and in some cases avatars that looked almost pornographic one user uploaded a photo of herself in a business suit the AI gives her back an avatar where she is wearing a bikini top another users a professional artist uploads family photos and the AI generates portraits where our chest is emphasized her clothing is revealing and the poses are suggested these women's didn't consent to this they uploaded normal photos the AI sexualized them without asking and here is the disturbing part this wasn't random this was how the AI was trained so stable diffusion the open source model that Lensa is built on was trained on billions of images scrapped from the internet and guess what people what's over represented on the internet sexualized image of women so when you feed the model a photo of women and ask it to generate artistic portraits it defaults to what it learned women should be posed in ways that are sexualized because that is what was in the training data I got them was an evil it was just optimized on bias data and nobody at prisma lab bothered to check or maybe they did check and didn't care because it wasn't slowing down growth I guess we will never know that the second problem was art theft artists started noticing something eerie they looked at lensa avatars and recognize their own style one digital artist saw an avatar that looked exactly like her work the brushstrokes the color palette the composition it was her style applied to someone else's face she hadn't consented to this she had never licensed her work to Lensa but her art along with millions of other artists work had been scrapped from the internet and fed into stable diffusion training data now this part a massive backlash artists started calling Lensa out they said she was making millions of dollars using our work and you didn't pay us you didn't ask permission you just took it Risma lab's response essentially this is how AI works we're not copying your art we are learning from it but to the artist this felt like theft what do you think because they had spent years developing their style and now an algorithm could replicate it in seconds and someone else was profiting from it the third problem you face belongs to them now here is the part that almost nobody noticed in the moment when you uploaded your selfies to Lensa you agreed to their terms of service and buried in the terms of service was a clause that said Risma lab now has a perpetual a revocable word by license to use your photo let me translate that you gave them your faces forever they can use it to train future models they can use it for marketing they can use it however they want and you can't take it back most users had no idea they just wanted cool avatars they didn't read the terms of service who does nobody reads the term of services but now millions of people's biometric data their faces are sitting on prisma labs server forever think about what that means your face is one of the most unique identifiers you have it's more personal than your social security number and you just gave it away for seven ninety nine and seven star ground cloud so by mid December of 2022 the backlash is starting to build artists are organizing they are creating hashtags no to AI art and art theft they are calling out Lensa on X on reddit on every plate form privacy advocates are raising alarms they are writing articles did you just give Lensa their right to your face while Lensa's term of service should terrify you women are sharing their sexualized avatars and saying this is not okay I did not consent to this and slowly the narrative starts to shift what was fun and viral a few weeks ago is now starting to feel exploitative take generally starts writing critical pieces NPR publishes an investigation into prisma labs history and finds that they have been involved in shady practices before failed apps and questionable data handling the European Union started asking questions gdpr Europe's strict data privacy law might not allow what Lensa is doing so regulators start investigating but here is the thing by the time the backlash reaches critical mass Lensa has already won they've already made in excess of seventy million dollars the founders have already extracted the wealth and they're a Russian company of creating an illegal craze of good luck regulating them and then in February of 2023 just three months after the explosion the data shows something interesting as revenue crashes downloads drop ninety percent the viral loops break people move on to the next thing the AI photo app trends fizzles now competitors enter the market the novelty we are soft and Lensa fades back into obscurity But Prisma Lab doesn't care. They got what they came for. They rode the wave, extracted the value and got out before regulations could catch them. So let me tell you what lens actually reveals. Because this story isn't just about one app. It's about a pattern, a system and a new way that wealth is extracted in the AI age. First is speed, beats, ethics. Lensa wasn't first. Avatar AI was first. Profile picture AI was first. But Lensa was faster at scaling, faster at viral engineering and faster at monetization. And in the time it took for anybody to ask, is this ethical? They had already made 70 million dollars. So this is the new playbook. Move fast, extract value and by the time regulators of critics catch up, you are already rich. Second was that open source AI is a double-edged sword. Stable diffusion is open source. That's beautiful in theory. It democratizes AI. Anyone can use it and anyone can build on it. But anyone includes bad actors. Anyone includes people who will use it to exploit, extract and profit without regard for consent or fairness. Lensa didn't build AI. It just wrapped it in pretty interface and a dark pattern paywall. And that was enough to make billions. Third is, virality is ingenial and not accidental. Lensa didn't go viral by accident. They studied what worked. They optimized the UI. They tightened the launch perfectly. They seeded influences. They build the shared loop into the product. This is what separates Figner from the first move was. Execution and speed. And understanding how human psychology works and exploiting it. Fourth is that nobody reads the term of service. You gave away your face. Millions of people gave away their faces for $7.99. And nobody realized what they were agreeing to until it was too late. This is the implicit contract of the modern internet. Convenience and exchange for your data, your privacy and your identity. And most people take the deal without even knowing they made it. Fifth, bias in AI training data creates real harm. The sexualization wasn't an accident. It was a direct result of training AI on biased internet data. And nobody at prisma labs cared enough to fix it. That is what happens when you optimize for speed and profit over safety and fairness. People get harmed and the harm is distributed unequally. Women, minorities and marginalized groups be at the brunt of algorithm bias. Six, the regulation is too slow. By the time regulators started investigating lensa, the moment was over. The app had already faded. The money had already been made. This is the fundamental problems with governing AI in the age of viral apps. The feedback loops are faster than governance. So by the time you write the rule, the system has already moved on. So let me connect back or connect this back to the bigger story I have been telling on this podcast. Lensa is the dark mirror of alpha goes move 37 that we distilled a few episodes prior to this. Move 37 was beautiful. It was strategic. It revealed something profound about AI capability. It was a moment where we saw intelligence that exceeded human intuition, but felt almost noble. Lensa is the opposite. It's not strategy. It's not noble. It's ruthlessly exploitative. It's AI used not to push the boundaries of human knowledge, but to extract wealth as fast as possible from people who don't understand what they are agreeing to. Both are systems that moved faster than human oversight. Both made the seasons technical business ethical to seasons at speeds that exceeded our ability to govern them. Alpha go move faster than lease at all could predict. Lensa moved faster than regulators could respond. One felt beautiful and one felt dirty, but both revealed the same truth. We have built systems that optimize so effectively, so quickly that human governance is structurally too slow to keep up. And the people who understand this, the people who know how to write the wave execute fast and extract value before anyone can stop them, they are winning. Now here is what terrifies me about Lensa. I would be honest. It's not that they did something illegal. I am not an illegal expert anyways, but if I had to bet, I would say that they probably didn't do anything illegal. Most of what they did exist in illegal creatures. And it's not that they were uniquely evil. They were not. And I know some of you would not like me saying that, but dozens of other companies are doing similar things. Lensa just did it better. What terrifies me is that this is the new normal. This is how wealth gets extracted in the AI age. You find an open source model. You wrap it in a viral product. You optimize for speed. You ignore ethics. I don't worry too much about it. And you extract as much value as possible in the shortest time possible. And then you move on. Lensa made 17 million dollars and 60 days. And then it faded. But the founders still have the money. They still own your biometric data. And they're probably already working on the next thing. And here is the scary part. Lensa is small. It's one app, one moment. Just fun viral way. Now imagine this pattern scaled across the entire economy. Imagine hiring algorithms that optimizes for profit over fairness, moving too fast for HR to audit. Imagine pricing algorithms that extract maximum value for consumers who don't understand dynamic pricing. Imagine credit algorithms that make seasons in milliseconds perpetuating bias at scale. Imagine identity system. You're in my world. Where our biometric data gets captured stored and monetized without meaningful consent. All buried in terms of service that nobody reads. Each one is a Lensa movement. Each one is a system moving faster than oversight. And each one is wealth extraction at speeds that regulations cannot match. And each one is accelerated. The question is not will more Lensas happen. They already are every day. The question is what do we do when exploitation becomes the business model and speed makes it unstoppable. So in December of 2022 millions of people uploaded their selfies to Lensa. They wanted cool avatars. They wanted to participate in the viral movement. What they didn't realize was that they were participating in something else. A wealth extraction machine that was moving faster than they could understand it. Nobody stopped to read the term of service. Nobody asked where is the data going. Nobody questioned why this free trial activity at 35 or 36 dollar per year subscription. They just did it because everyone else was doing it because it was viral because it felt fun. And by the time anyone realized what was happened, Prisma Lab had already made their money and moved on. So here is what I want you to take from this story. Speed is not the same as safety. Liberal is not the same as ethical and convenience is not free. You are paying with something. You just might not realize what it is until later. I want you to notice the next Lensa moment in your life. The next app that goes viral. The next product that feels too good to be true. The next time you are asked to upload something personal, your face, your data, your identity in exchange for something fun. or read the terms ask who benefits from this what am I actually agreeing to and is the trade worth it because lense was just the beginning this pattern is going to repeat and it's going to accelerate and the only difference we have is awareness. Not regulation because that's too slow not hoping companies will do the right thing sometimes so we won't because speed and ethic can beat ethics and if you want to it can beat it every time so that leaves you the last and the only line of defense asking questions demanding transparency and saying no when the trade isn't fair. Thank you for listening and for spending time with this story. This is Rohit your identity navigator. If you want to go deeper if you want to understand how identity data and AI are colliding in a ways that most people don't see subscribe to the podcast you can also find me on my sub stack by identity navigator. I'm writing about these moments I'm talking about these patterns that systems that are moving faster than we can go on then and if this episode resonated with you I would request you to like it share it because these stories needs to be told. These warning needs to be heard until next time stay aware and protect your identity because in the AI age it is the most valuable thing that you own. Thank you for listening this is Rohit your identity navigator.