← The Digital Executive
The Digital Executive
Building Digital Trust: Kaarel Kotkas on the Future of Identity Verification | Ep 1142
00:15:36
Loading summary
Transcript13 lines
- [00:00]
A
Foreign welcome to Coruscant Technologies, home of the Digital Executive Podcast. Do you work in emerging tech? Working on something innovative? Maybe an entrepreneur? Apply to be a guest at www.corazon.com brand. Welcome to the Digital Executive. Today's guest is Karel Kocas. Karel Kocas is the founder and CEO of Verif, a global identity verification company trusted by leading companies including Uber and Bumble. As a visionary and strategic thinker, Carrel drives Verif's mission to stay one step ahead of fraud and competition in the rapidly evolving digital world. Known for his energy and enthusiasm, he inspires the Verif team to champion honesty and integrity online. Carrel's leadership has earned him numerous accolades. In 2023, he was named to Forbes EU 30 under 30 and in 2020 he was EY Entrepreneur of the Year in Estonia and the Nordic Business Report recognized him as one of the 25 most influential young entrepreneurs in Northern Europe. Well, good afternoon Karel. Welcome to the show.
- [01:14]
B
Thank you for the invitation. Great to be here.
- [01:17]
A
Absolutely, my friend. I appreciate it. You're making the time. You're about a seven hour difference from where I'm at. You're in. Actually, you're probably more like closer to eight from Estonia. I'm in Kansas City, the central U.S. so again, appreciate you making the time. It's hard to do these sometimes, navigating the world's time zone. So jumping in. Growing up on the Estonian island of Humana Huma, I'm trying to get that pronounced correctly. You experience firsthand how fragile online identity systems can can be. You've told that story of editing your ID to make a purchase, right? How did that childhood moment seed the idea for Verif? And how did you translate that personal insight into a scalable business model?
- [02:02]
B
I grew up on a farm and on this small island in Estonia. Then as I the youngest one in my family, then I was automatically the IT guy at home and all the online orders and everything that I needed to do showed me that when I was asked to upload my ID and it wasn't accepted because I wasn't 18 years old then as git, it felt like an instruction. So I just changed my date of birth and everything and all the orders got confirmed. But this wasn't exactly the time that I felt this is becoming my life's job. This is something that I'll do now. It was many years later moved to Tallinn and when I was about to graduate my secondary school, then in the last grade in the high school, I was asked by Wise to Test out their identity verification solution and see if I can find some ways of improvements. But this was eventually the time that it got back to me. Let me just test out a trick that I did when I was a kid. And then I found out that back in 2015 the similar issues were still present. It was global identity verification market that was driven by the compliance demand and people still looked only on couple of pictures where they did as much as needed and as less as possible to an extent. Yet the cost of doing fraud on pictures was so, so low and it was too easy to be to pretend to be somebody else online that you're not. So this was something that got to my head by thinking instead of like three pictures, let's have video from the beginning till the end. Let's have device and network fingerprinting at the verification step, let's leverage the behavioral insights and eventually become the registry on our own. And this is how what if got born. On the simple belief that leveraging over thousands of data points per session enables you to give so much more accurate decisions compared to three pictures like selfie and document front and back alone. And as a result this is how verified now already 10 years ago.
- [04:36]
A
That's amazing. Thank you for sharing your story too. I think that's important how you got here, right? That story as a kid helping you were like the tech support for your parents and you found out quickly that you can kind of fake a system. But over time that came back when you were asked to do some testing in the identity verification space. But you're absolutely right, we need to have more fingerprinting at again device application, wherever that is, leveraging all those data points through the process to validate this stuff. I think that's really important, especially in today's age. So thank you. And Carl, you said that you believe online identity verification can be more secure than face to face verification. What does secure mean in this context? And how do you balance speed, usability, compliance and anti fraud rigorous in these systems that you're building?
- [05:26]
B
Coming to the first part of your question, like how we can make online identity verification more accurate and face to face verification then we as humans we are subjective. We are subjective by nature and there is very easy for us also to actually pretend to be someone else or outsmart other human beings. So this is the first part where I've seen too many times the cases where like when we are taking United States as an example, if I'm asking from the crowd, please do raise your hand who had fake ID. You allow yourself appear before you were 21 years old, then it takes time to see the first and then later there are very many which are raised. And what is the key topic there is that it's so much easier to outsmart a pouncer or someone on just a plastic paper that looks okay, you're close enough. Go ahead. Compared to having a session that leverages and makes a decision objectively by having very many data points and is solely focused on driving the most accurate and scalable verification result as the more information points you have, the more highly you can automate it with accuracy. And as a result you can build so much more accurate decisions which are objective then face to face authentication as a result. And if I take let's say a real world example that let's say there's a person going to a bank with someone else's ID and if the banker doesn't believe that this person truly tries to be someone else, that they truly are, then this fraudster in many levels need to rely on their fast legs and run away. There's not as much that will lead to conviction in this particular case. But on the Internet we are still facing a challenge with accountability. And let's say I wouldn't want to be the one pretending to be, let's say you behind my personal computer at my home as then we'll have also much more insights and information that will lead to conviction than physical interactions.
- [08:03]
A
Although thank you, appreciate that. I like how you broke apart. Nowadays the physical ID you can still actually fake someone out or fake a person with a fake id. You brought up the bouncer or the bank example. But with digital ID there's a lot more information points. There's more accurate decision making in this verification process and I think that's so important regardless of what types of transactions we're doing, whether it's online or in person. I think that's important. So thank you. And Carl, the next question I have for you. Verif uses automated decisioning, behavioral data, device context signals and human review when necessary. How do you decide what gets automated versus what must remain? Human review, especially as synthetic identities and deepfakes become more convincing.
- [08:49]
B
There's always the task. Then deterministic and probabilistic models are better fit for the task at hand. As machine learning is deterministic and the latest in gen AI and foundation models are more probabilistic. And if I think along the lines we as humans, we are also probabilistic more than deterministic. So it's always leveraging the right tool for the task at hand from the toolbox. And when we are doing millions of verifications every single day, then we are also able to roll out more and more new rules that enable us to find out what is something that is out of ordinary and requires very thorough review to prevent new attacks as the fraud is always evolving. And the accuracy of the new models, they all depend on the accuracy of data it's been trained upon. So this is where we as humans are still better on really detecting from the bias data set additional patterns that we need to review which will become part of the next model training to capture the new upcoming fraud trends at scale. So this is an area where we as humans here we are making, let's say mistakes on very repetitive tasks where computer is better. But when it becomes more creative as fraudsters us, we also need to have great events, a little bit like a hacker mindset as humans to really pick the similar patterns that enable us now to build new automations upon. And that's the reason why we as humans are going to be essential forever as long as we are verifying humans.
- [11:05]
A
Thank you, appreciate that. I've done a lot of podcasts on AI and some digital identity and there's certainly a lot to be said about that. But you did talk about deterministic and probabilistic models, right to do this work. But what I really like is this human in the loop. We've got to be creative, we've got to have that hacker mindset as humans to meet the challenges of these fraudsters. So I think that was important. And Carl, your mission and Verif's mission are about creating digital global digital trust so services can be accessible to all, everywhere. Looking ahead maybe five, 10 years, what is your vision for Verif and the broader identity ecosystem and how do you hope your work contributes to a safer, more inclusive digital economy?
- [11:52]
B
It's interesting that today the best way for verifying people's identities are our government issued IDs. But when I'm also thinking along the lines what does the government issued ID mean for the people the moment they are born, then for the first 18 years it mainly thinks that you are someone's son or daughter born in this country. And as there is not so much interactions that we can have at the beginning of our journey and even when we are going ahead, then government issued ID is still not meeting the complexity that our real identities are. And it's still quite static in nature with its power. And even at times when people say all the passports are equal, then we know some passports are more equal than the others even when we are taking just traveling as a easy example. So what I'm thinking of is the part where the tech revolution and the Internet economy which has enabled us to access the whole world from distance, it also has been the key as driver for globalization. Yet our identities are still covered with the borders. So if I'm thinking for the next five to 10 years then I'd like deriv to build towards the world where we come together as an industry, where we drive collaboration as an industry against fraud, but also move towards least friction for the honest and, and for the honest people. I mean I really want to make sure that we can leverage and build up our real identity Coming from the Barinhoff crust as when we've been good actors across our Internet interactions then this should be building up the pattern of trust. And the pattern of trust also becomes so much stronger source of our true identities than government issued idea can be though. My dream is to build towards the world where let's say passports will be issued by veri, not only by the governments, to truly enable people equal access to services, location independently, no matter, no matter where they from, when we can be the source of truth for trusted identities, whatever.
- [14:43]
A
Awesome. Thank you so much. And there's so much to be had with a government id. You walked us through that whole thing just not meeting the needs of digital identity today. And as you mentioned, it's a lot more to do a passport, there's a lot more to borders and traditional verification. But I like your vision, this idea of a pattern of trust where we can be more digital inclusive, where it's equal access to services as you mentioned. I know there's this idea in the cyber world, cybersecurity, zero trust. But I like how you talked about that pattern of trust and I think that's really important. So thank you and Carl, it was such a pleasure having you on today and I look forward to speaking with you real soon.
- [15:26]
B
Thank you Brian, it was great.
- [15:27]
A
Bye for now.