What's Up with Tech?

When AI Meets Identity Protection: The Full Story of Loti's Internet Takedown Service

Evan Kirstel

Interested in being a guest? Email us at admin@evankirstel.com

Imagine having a delete button for the internet. That's exactly what Loti offers—a groundbreaking service that literally downloads the entire web and runs sophisticated face and voice recognition across billions of pieces of content to find and remove unwanted digital traces.

Luke Arrigoni, Loti's founder, shares how the service evolved from combating non-consensual intimate imagery into a comprehensive digital protection platform now backed by major entertainment agencies and venture capital. What sets Loti apart is their extraordinary technical approach—they can identify as little as 15% of a face or voice, making them uniquely effective at finding content that's deliberately hidden or manipulated. The company processes takedown requests within 24 hours, dramatically faster than traditional DMCA requests that can take weeks.

Corporate executives have become major users of the platform as deepfake threats proliferate. Scammers create convincing impersonations of C-suite leaders to sell fake access to companies or release fraudulent financial information. Meanwhile, celebrities and public figures use Loti in surprisingly strategic ways. Rather than removing everything, most clients establish specific rules about what should remain online versus what should be taken down, understanding that their digital presence forms an important part of their public identity.

The conversation explores fascinating ethical dimensions of digital protection. Loti operates as a "strategy engine, not a morality engine," allowing clients to set their own boundaries. While they support child protection organizations and partner with major social platforms to streamline takedowns, they deliberately avoid providing "surveillance as a service" to law enforcement, recognizing the complex ethical questions such arrangements would raise.

Ready to protect your digital self? Visit Loti's website to explore their services, including a free option providing five takedowns monthly for individual users who need protection without the hefty price tag.

Support the show

More at https://linktr.ee/EvanKirstel

Speaker 1:

Hey everybody Talking about protecting your digital self online a fascinating topic around digital rights protection with Loti Luke.

Speaker 2:

How are you? I'm doing well. Thanks for having me, Evan.

Speaker 1:

Thanks for being here. Really intriguing topic, very timely. Maybe start with the big picture introductions to yourself and what inspired the creation of Loti, with the big picture introductions to yourself and what inspired the creation of Loti.

Speaker 2:

Yeah, I guess the big picture is we like to download the whole internet and then we run face and voice recognition over all of it Wow and we try to find our clients in that. And then they have all these rules around what we can take down. So they'll say, hey, if there's anything that's an impersonation or explicit content or a deep fake, just issue a takedown on that. So we do that. So think about us as, like this huge search, find and destroy cluster, we basically give delete for the internet, for people.

Speaker 1:

I could use that. So what inspired the creation of Loti? What were you aiming for when you started out a number of years back? So I used to do a number of years back.

Speaker 2:

So I used to do a lot of consulting and there was a client that wanted partial face face recognition so think like 15% of the face and we took this tech and we kept intellectual property from the project and then we pointed it towards the internet because the thinking really was I think I was watching euphoria or something and there was this big problem with non-consensual intimate images. So colloquially everyone calls it revenge porn, right, but there's this idea that there's really nefarious sites where people upload content and it's hard to find everything and it's hard to delete it, and this would be the tech that could make that happen. So we just created this little service and we pointed it out at the internet and we allowed a lot of people to come through and just find stuff for cheap or free and remove things. And then, in 2023, sag-aftra went on strike and I used to work at entertainment.

Speaker 2:

I used to work at Creative Artist Agency and I said, hey, I have this service. I've been running kind of as like a goodwill thing for the Internet but like we could expand it pretty massively and solve this problem for everyone at a massive scale, Would you be interested? And the agencies were like, yeah, that's, this was no brainer, let's do it. And so it's been kind of a wild ride since then, because the capital that we've got from a seed round last summer and then a series A a couple of months ago from COSLA, it's just been able to let us expand this thinking around. How do we search and find things across the internet that are hard to find deliberately hard to find and then how do we use our influence to take those things down rapidly? The whole target at our company is less than a day and yeah, it's been such a crazy ride actually.

Speaker 1:

Well, it sounds like it and such an important mission. So I imagine you have many different types of customers and prospects consumers but what does it look like in practice for enterprise users or business users?

Speaker 2:

A lot of executives get impersonated. So it's common, for if you're a publicly traded company, people have motives to create deep fakes of you saying things Think about like a fake earnings call, right. But most commonly it's executives or ancillary executives that are selling access to the company. Now, it's not a real executive doing that, to be very clear. It's someone pretending to be the COO of a larger company and saying you know, hey, we're trying to do some sales pitch whatever and for $100 you can, you know, in crypto, I'll give you access to the CEO. It's kind of bizarre, but these kinds of scams are pretty prevalent. So there's that, and there's also just intellectual property protection. So making sure Buzz.

Speaker 1:

Lightyear is protected as well. So obviously you're a heavy user enabler with AI right now and agent-based tools and other things. But how does your approach work exactly? How does it differ from other tools or alternatives that might be out there already?

Speaker 2:

Yeah, the, the. The real competitive advantage is we. We have depth and breadth. So we have, like, we have, uh, we we're ingesting the whole internet, which is a wild and expensive tech, right Like it is. It's very, very expensive and expensive tech, right Like it is. It's very, very expensive. And so that's our differentiator, for one. But also the depth is is there in terms of we, we could do partial faces, we could do voices. There was a lot of ways we even can detect animations Like so we know what, like, Mickey mouse looks like, right, and so this is in this respect, being able to ingest so much and have so much fidelity on that and at the same time, because usually you do a trade-off of one to the other, is something that has really put us in a league of our own.

Speaker 1:

Wow, fantastic. And you know what does it mean when customers ask you about trust and safety and controls I mean the enterprise. They have a lot of these kinds of requirements. How do you work with businesses?

Speaker 2:

Yeah, so there's some surface level answers. I can tell you, like SOC 2, right, but there's like or for anyone that's not familiar with SOC 2, there's like security compliances that a lot of enterprises want everyone, even their smallest vendors, to be compliant with, and we understand and comply with those, especially in this space. When you talk about, you know, privacy and security, our system has, you know, 10 billion pieces of biometric material, right, faces and voices. Like we're tracking most of the internet, like it's really, really important for us to make sure our systems are secure. So, as a as a funny bit, like our security team, we have an offense and defense security team and some people on the offense side are constantly trying to hack our own system and the defense side play against it. That's not suck too. That's like above and beyond, right, but we're really, really interested in having a very secure system. And then the last way I think I would answer the question is people are curious about what it means to be trusted on the internet, right? Like how do like? There's you, the real Evan, and there's a fake Evan imposter out there.

Speaker 2:

Sometimes the conversation around trust and privacy and things of that nature really start to dive into. How do I make sure that my fans trust the real me, right? And that's a bigger question. Obviously I'm biased. I have a great service that will help people do that. But there's a lot of ways that people try to tackle that problem Right, and it's becoming a growing problem on the Internet of you. Know, if you, there's a funnier movie movie. It's not really a comedy, but it came out mountain head uh hbo last week and I've heard so much about it. But really, like there's a growing concern that everyone has, both in in in art like that and just in in a normal conversation like this, where we wonder, uh, what does it? What will the internet look like when we are in some kind of post-truth era? And I don't have great answers for that.

Speaker 1:

Just the good questions. So you must see lots of really unexpected, interesting ways customers are using your platform. Care to share any stories or anecdotes in that regard.

Speaker 2:

Yeah, we tell people we run a strategy engine, not a morality engine. So for certain celebrities, we don't care what the rules are that they create of what they want to take down. So they might say you know what? I don't really care about any of the explicit content, you can just leave it up. I really care about just imposters. We are not going to tell them. Oh no, it's important for the Internet that you know we don't care, and so it's really funny when you think about that. You can make whatever rules you want and you understand how fame and artwork and the combination of the two and these people that are public figures. They deeply understand their audience, they understand the power that they get through the like, the podium that they have, and they're very mindful about what they take down. You would think that, given the ability to find everything and delete it, a lot of people would do that, but in fact that is not the case. In fact, they're very strategic about the areas in which they would like to be represented or not represented online.

Speaker 1:

Interesting and you're a global platform, I imagine. Considering the internet is global. How does that work? In different regions and countries, different rules, regulations? Yeah, I assume you're not just a U S service.

Speaker 2:

Yeah, you know, we, um, we spend a good portion of our sales process telling people what we don't do.

Speaker 2:

So we aren't effective on tour, like, which is like the dark web, the, the black market of the internet, right, uh, because if we issue a takedown there, they have no reason to listen to us, right? So, uh, we make sure everyone understands, like, by the way, I'm not going to be able to do anything there and we don't work on in russia. Russia has rebuilt their entire financial system and everything else out of um, out of everything that's come out of that conflict. So we are also really straightforward that, um, you know, hey, if there's a site in Russia and it's fully Russian, like Russian domain, russian servers, um, it's basically us asking, nicely, they don't really have any reason at all to take things down. And so, because of these two areas that are, you know, and I would, I know, tort isn't technically talk to globalization, like the question you asked, but in some level it is a little bit which is, you know, there's some areas of the world, both physical and theoretical, that we cannot touch.

Speaker 1:

I bet. And so how do you tie into what's happening on the regulatory legal side? You obviously have big legislative requirements DMCA and all kinds of related rules. How do you keep on top of everything that's happening there, or does that sort of play a part in?

Speaker 2:

your work. Yeah, we've been really lucky that, for at least in the US, we've been asked for guidance on all of those rules. So the laws that are coming out like take it down and no fakes. There's a lot of really smart people that are pushing that forward, that are authoring that, and we've been lucky enough to grab Zoom calls with them to tell them, hey, this is the thing that we need on the takedown side, and they're very nice and they all are like, okay, that's cool. And so what we care about is what is the language for takedown? Because if you make it like DMCA language, then you get two weeks to take things down which is in 1999, like two weeks on the internet probably felt like the right amount of time Right now.

Speaker 2:

Two weeks on the internet, that's an entire, that's like four press cycles, right, like that's something that's gotten big and then shredded and then going up. That's way too long, it's an eternity. The internet history of the last two weeks has been crazy, right, and so, anyway. So the language we've asked for is like hey, can we get something that's more compressed, like 48 hours tends to be the thing that we, um, we've been trying to push there and then, at the same time so, while that's a tailwind for us, like having a lot of the the us legislation say, if you are being deep faked online and it's you did not authorize it, you have the ability to issue a takedown. That's a wonderful gift for our business, right, uh? And then you have other jurisdictions that are trying to say, you know, like no face recognition at all, like carte blanche, not, not nothing at all, like full stop, no face recognition, and we have to go to those jurisdictions and tell them I apologize, but this is the only way to solve this crisis, right, like I, if I can't, if I can't find you, right, if I can't identify that it is, evan.

Speaker 2:

It doesn't matter if it's a deep fake or not, right? There's lots of things that are deep fakes. In fact, people will use filters to make themselves look prettier better. Put cat ears on themselves, right? That's technically all AI. That's all technically a deep fake, right? What really matters is the content. Like what is it and is it you? Everything else just immediately becomes immaterial, and so, when it comes to legislation, we're getting some really good stuff and some other stuff we just have to be more educational to those legislators.

Speaker 1:

Interesting and tell us about your go-to-market today and maybe in the future. Are you going directly to these customers? Are you partnering with any of the big tech players? How might that look over time?

Speaker 2:

So right now, go-to-market is all of our public figures. We use distribution channels. You can't really cold call an ex-president right, that's not how it works right.

Speaker 2:

But, um, we know the markets or we know the channels that they'll go to to seek help on this particular product or this particular problem. And, uh, we go to those places for consumers. Uh, the product is really straightforward you go to the site and sign up. Right now we don't really have. We don't spend a lot of time or money or ads on that because it's almost free. In fact, there is a free version where you get five takedowns a month. So we don't really have. We don't spend a lot of time or money or ads on that because it's almost free. In fact, there is a free version where you get five takedowns a month. So we don't really make money on that, and so we're not trying to like push it wildly.

Speaker 2:

It's just, if you have this problem, we're already running this cluster. We're solving it for a lot of public figures. You can just come and sign up and we'll take it down for you too. We're trying to be a good service to the internet and trying not to be a war profiteer, like if you have a particularly bad problem on the internet and you are a consumer and you need five things taken down, it feels odd for us to really gouge you for that.

Speaker 2:

But you mentioned tech partners. Yes, we are partnering with some larger tech, like one of the largest larger tech, like one of the largest several large platforms for social media, but it's not for them to do compute or to use their technology. It's actually because they want to be really good partners and they want to have things come down faster as well. So they'll partner with us and they'll say, hey, lodi, can you send us the list of things that need to come down this week? And we'll say absolutely, and we'll send it off to them. So the partnership isn't really even financial. It's literally hey, let me take care of this problem with you, and we've been lucky enough to find the right people at all those companies to push that.

Speaker 1:

Fantastic. You must be an interesting partner to law enforcement agencies. So we have huge problems with revenge porn and child exploitation and all kinds of other things. Is that a market for you or an opportunity, or just gratifying that you can?

Speaker 2:

help. So, when it comes to CSAM, like child exploitation, we're really trying to work with NCMEC National Center for Missing and Exploited Children and not really for profit either. They don't have the budgets they needed and that feels we really trying to work with Nick Mick, like national center for missing and exploited children, Uh, and not really for profit either. Uh, they don't have the budgets they needed and that feels really odd for us to try to charge money on that front. Um, so we, we are trying to work with the right agencies to use our superpower for good. Um, when it comes to law enforcement, though, you, we are shying away from that.

Speaker 1:

It's.

Speaker 2:

A lot of it comes down to to be able to give you know. People jokingly call our service like God mode for the Internet and to give that to every law enforcement. It opens a lot of ethical questions and, in some jurisdictions, a lot of legal questions, and so right now we are we are actively avoiding that quagmire and so we are not selling it to any law enforcement. We're not selling surveillance as a service. Uh, it is really just you being able to find you online and that that's certain. That, like, uh, like, all the plants align. For that the ethos is spot on. It's hard to find anyone that would disagree with that kind of service. Right, like being able to find you and delete you on the internet. Um, it's very there's. No, I don't have to think about the ethics behind that.

Speaker 2:

With law enforcement, though, it's a little bit more dicey. You really have to consider when is it appropriate, when is it not? You mentioned one example. Like what, if there's a missing kid? I probably could get 99 out of 100 people to say, yeah, go for it, right, and then from there all the way forward to hey, I have an idea that someone may have committed a crime. Can I put them into our system to see if they've committed. They were anywhere in the area at the time like deeply circumstantial, trying to strengthen a speculative argument, which you know that's a lot harder for us to participate in. I could get maybe one in 100 people to be like yeah, that's OK, you can let people just randomly try and make their own cases, so we're avoiding surveillance as a service.

Speaker 1:

Got it Well. I'm sure the three-letter agencies and others have their own version of that that they've created A peek behind the curtain. What does your technology infrastructure look like? Are you running your own data centers and GPUs and LLMs? How does that look?

Speaker 2:

We're in every major cloud provider. So right now we aren't running our own data center. We're still a Series A company. We're still a smaller company that is likely on the roadmap. We will be building a data center at some point. But right now we've deployed into every major cloud provider infrastructure provider because we do consume a lot of GPUs. We basically boot up most of the GPUs on the West Coast for AWS at night and eat all of the Internet and then turn them off. So there's kind of a crazy amount of coordination and infrastructure and compute that all have to kind of come together to build the product we have. Yeah, Amazing.

Speaker 1:

And all of these enablers, new language models coming out, new hardware versions and tools, gifts for us.

Speaker 2:

How will they?

Speaker 1:

enable the business.

Speaker 2:

You must be a kid in a candy store in terms of all of these capabilities out there every day, every week I've been saying it more and more often lately but like, uh, it feels like we just win lottery tickets after lottery tickets at lodi, uh, and really our job has become being just like a good steward of that right, like we have to be like a good steward of the lottery ticket. We have to make sure we go to the, the store and cash it, and so part of that is also, yes, we're being given these models that are magical for us in terms of understanding language. We wrote all of our visual models. We wrote our audio models. It was like understanding who you are, like the identity portion.

Speaker 2:

There's a lot of custom. There's like 40 custom models we built, but there's a number of open source models that have been enormous gifts to our company, and I would, I would, I would anyone that's listening, I would say don't shy away from that right. Like, if you have access to the open source stuff you don't, it's not a it's not bad to use their stuff, even if you are building your own custom cool things. There's no reason to turn down any tool that'll help you succeed.

Speaker 1:

Great advice. So what's next, I dare to ask. You're doing so much today. What are you excited about over the next year or two?

Speaker 2:

It really is an expansion of our defensive product, right? So we want to make sure that things that are generated are generated with the right rule sets, and so when we see things that get generated and they get put out on the Internet somewhere, we want to make sure that hey, that matches up. You're good to go. You can make that of that person Kind of back to something we talked about at the start, which is public. Figures tend to have very strategic ideas around how they're represented and in a Gen AI world, they're not going to want to not be generated, right. They'll want to be generated just to follow the right rules, and so there's an expansion of that thinking that we're going to do in the next year Fantastic.

Speaker 1:

So interesting. Yes, can't wait to see all the good stuff ahead, onwards and upwards. What a great service. I can't wait to try it out personally on a few things, and I'm sure many of our listeners would love to put this in action. Thanks for joining, luke.

Speaker 2:

Of course, thanks for having me.

Speaker 1:

All right. Thanks everyone for listening and watching and check out our new TV show, tech impacttv now on Fox, business and Bloomberg TV. Take care.