AI’s Influence on Elections – Techstrong AI Podcast EP37

Hello, and welcome to the Techstrong AI Podcast. I'm Amanda Razani, and with me today I'm excited to have the Chief Technology and product officer of Julio, which is Bala Kumar. How are you doing today?

Hello, Amanda. Thank you for having me on. How are you doing?

Happy to have you on our show. Can you talk a little bit about ju o and what services do you provide? Absolutely.

Jui is everything related to government issued IDs. So if you think about a driver's license and a passport, we are the ones who provide the verification and authentication of those, um, those, uh, form factors in the digital world. So think about going through airport security.

You produce a government issued ID to make your way through the security lines. It's the same type of service except that's done online in the digital world. Okay, wonderful.

Well, that brings us to our topic of the day, which is ai, DeepFakes, and, and security. So you recently put out a survey and I want you to share a little bit about, um, the survey and who you pulled and everything, but it, it talked about the influence of DeepFakes and AI on the elections. So can you share a little bit about that?

'cause that seems really concerning. It is. Um, everyone's talking about AI recently, especially Gen AI with chat GPD and other types of capabilities that are now available to, uh, common folks on the ground.

Um, even the person has access to these types of tools now, um, which means, uh, a lot of folks, uh, online, especially fraudsters, also have access to the, to this very sophisticated technology. And so what we are seeing, um, in our, uh, environments is a significant increase in, uh, uh, fraudsters using these technologies to generate deep fake images, um, and send them through, um, posting like they are real individuals. And so, not only do we see this, uh, in, in our world, this is something that, uh, we are also seeing out there.

Um, 72% of, uh, American consumers are very much worried and concerned about the potential for AI and deep fakes to influence the elections that are coming up here. I think we are about 40, 40 ish days away from, uh, presidential elections. Um, a significant, uh, percentage of our folks, 72% of folks are very concerned about the use of these types of, uh, technologies for nefarious purposes.

Um, and they do believe, uh, that these deep fakes will undermine our trust in politicians and in the media because you can use these deep fakes to, uh, create voice recordings, um, that are post, that are made look like it's coming from some of our leaders. Um, they're able to super and post, um, pictures with the faces of our political leaders and make it look like they're the ones seeing stuff that they are not saying. Um, and so this can cause a lot of, um, confusion out there.

It can cause a lot of misunderstandings. Um, and that can have significant impact on the outcome of the, uh, re elections itself because if, um, if you add this deep fakes, uh, sending out messaging from candidates that they do not represent, but that's essentially what people are hearing, they're gonna believe that's what the candidates stand for. Um, and so there's significant increase in skepticism in the content that folks are seeing online.

So I mentioned about 72, 70 5% of the consumers are concerned about the elections. Um, there's about 70%, um, folks concerned about the content that they see online compared to just the last election. So we are talking about just the last three and a half years or so, just compared to the last election.

There's, uh, a, a significant, uh, significant population that's concerned about it. Um, and the other interesting aspect here is about half the population, about 46% of global consumers, not just here in the us. Uh, 46% of global consumers believe that they can easily spot a deep fake of a political figure or a celebrity.

Uh, when you think about it in US terms, that drops to about 37%, and that also changes by age group. But, but the problem here is a lot of people believe that they can spot a deep fake, but in reality, when you actually put them through the test, they cannot. Uh, we just hosted a event in, um, in London a couple weeks ago.

We are on the ground with a bunch of our customers, and we had a few, uh, pictures on the, on the screen with, uh, real and fake images. Not one person could identify, which was got it right when we asked whether it was real or fake. Um, so that's a significant concern, right?

So people have the assumption that they can spot on deep fake, but in reality they really cannot. And so, um, that also makes it difficult for them to figure out what is good content or what is fake content. So a combination of these is what's, uh, causing a lot of, um, skepticism out there for, um, uh, for consumers.

That's a very serious problem. I mean, we already have many skeptics here in the US about, um, the election, the last election. So now you've added AI and deep fakes and these robocalls, now we have even more skepticism.

Uh, so is that gonna, you wonder if that's gonna keep people from voting because they don't trust the system. And we've seen, we've seen this in other countries as well, where AI was, um, used to disrupt, um, the election process in other countries recently too. So if, if they don't have a way of recognizing deep fakes, even though they think they do, my question is, how do we solve this problem?

And do you have any tips for people to recognize DeepFakes? There's really not too many tips, uh, in terms of our ability to detect deep fakes. Um, because like I said, even good images sometimes look like it's manufactured, manufactured images look too good.

Um, and, um, it, it, it is hard for the human eye to be able to catch this. I have, uh, I've been in the space for a decade plus, and I see images come up and I'm not sure if it's real or fake, right? There are some that are obvious and you can catch them, but some of the more recent ones that we are seeing, it is really hard for us to detect it.

So, um, I don't have too many tips, uh, around this, but what I would say is just being better educated, um, knowing that this is a possibility, um, going try, try and go to the source of the data rather than just what you read on social media or what you hear on social media, um, is what I would advocate for. Um, because if you go to the candidate's website, um, if you, uh, go to the news media, um, and, uh, listen to, uh, what the candidates are saying that is more closer to the truth than what you consume, uh, on any normal websites or any social media posts that people are putting out there. So those are the ones that I would avoid.

Uh, it's okay. I mean, it, you cannot shut down social media, of course, so folks can, uh, look at that information. But if there's, if they see something that does not line with, um, the political party or the candidate, et cetera, please go, uh, to the source, um, and validate, uh, what you're hearing.

com tweets. So right now we're talking more about the manipulation and the misinformation, but, um, are there threat actors there that, uh, we need to be concerned about that will use AI to actually impact the polls, the outcome of the polls? I do not believe so because the, uh, election missionary, the voting missionary has, has been time tested.

Um, and over the years we have seen that, yes, there are very small instances of fraud, but these are really outliers. Uh, but for the most part, uh, I think the, I believe the government has got the right controls in place. Um, and so when somebody actually gets to the polling booth and they're going through the, the, the process itself, I think I believe that the, uh, government's got the right controls in place.

So I'm not too concerned about that. Uh, for me, the bigger concern is really around misinformation, uh, that can be floating out there and generated content that is not really coming from the candidates that can cause confusion for, uh, folks out there. So when it pertains to the enterprise, do you have any, um, tips or advice for business leaders when it comes to how they think, uh, and protect against ai?

Absolutely. I, and this is what we at Jui do on a everyday basis, and not just ju o companies in this space. This is essentially what we do.

Um, the, the good news is we saw this coming, so this did not catch us by surprise, uh, and because we were expecting it, we were already, um, working on how we would respond to this type of a situation. So over the last several months, a year and a half, plus, we have already been investing in this area. Uh, we believe you can fight gene AI with gene ai.

You can fight AI with ai. Um, let's not forget that these tools are available to us, but the same tools are available to fraud as well. And, uh, let's not, uh, underestimate their ability, uh, to mount some serious, uh, attacks.

So when this technology started to come out, we quickly jumped on it, we started to look at it, and we started to incorporate that already into our solutions. And so the good news is, um, as this is now starting to become, uh, mainstream attacks that fraudster are using, um, it's critical for enterprises and organizations to start using AI based technology to counter to fraud. And so, couple things, right?

So if you are, um, if you're trying to verify a driver's license or a passport online, this, the, the good news is there's a lot of security features that's embedded within the ID itself. So using AI technologies to look at those, those intricacies of those, uh, securities that are embedded into those documents, like the holograms and the watermarks, et cetera, uh, these are, these are difficult to replicate. Not impossible, but it is not straightforward.

Um, and then the second aspect is being able to retrieve information from, uh, those documents and validating it against other sources as well. So as a cross reference of that information across other sources, so you can improve the accuracy of, uh, verification. And then it's not just about the ID itself, it's who's presenting that id.

Um, so if I'm the one who's submitting the id get a selfie. And remember, good uses will happily provide a selfie for a fraudster. A selfie is the equivalent of a mugshot, right?

And so there's only so many times they can use the same phase as they go through this process. And if they have to use the same phrase, face across different IDs, now they have to take that face and also embed that into each of these different IDs, right? So it'll have the same face, but different names and different addresses, different document numbers, different document types.

And so there are ways for us to be able to detect these types of manipulations and call it out. Um, and so you may miss one or two, but you're not gonna be miss missing a pattern of these, right? And so it's critical for organizations to use, um, good AI techniques to combat, uh, fraudulent IDs include selfie capture, and it's not just a selfie.

Make sure that you're also doing what we call a liveness check. So you're not holding a picture up to the camera. You're actually a person with, with facial movements, right?

Your lips are moving. Um, and, and so when you look for that liveness, you know, it's an actual living, breathing human being that is actually interacting with the camera as it taking that selfie and then comparing that selfie with the image that's on the ID as well. Um, and there's also ghost images on the id.

So there's multiple ways in which you can do this. Uh, but what is critical is to take a layered approach, not, it's not gonna be one size fits all. It's not one tool that's gonna fix it all.

So using the, uh, the good tech good techniques to look at the security controls on the id, uh, taking a selfie, making sure that there's a liveness component that's tied to it, extracting the information from the ID and cross-referencing it, uh, which is what device is being used for the transaction. Is it a cell phone? Is it a mo, is it a iPad, is it a laptop?

Is it the same device that's showing up multiple times, right? When you start looking for these types of correlations, you will have a really good robust solution in place. And so that's essentially what we strongly advocate organizations to look for.

And over a period of time, you can also start incorporating things like behavioral movements, et cetera, because each of us have a very different behavioral signature that's tied to us. And so those are additional aspects that, uh, we strongly encourage organizations to uh, consider. Um, and like I said, this is what we at Junior do, uh, for a living and not just ju uh, other companies in this space.

This is basically what we do. What's most important is how can we all provide the right services to enterprises so we can safeguard the common folks out there and not let them become victims of identity theft or to, uh, allow for creation of synthetic IDs using, um, that information. Absolutely.

And threat act threat actors are getting more and more smart with their tactics and, and there's all these tools that are advancing rapidly. So what is the key takeaway you would leave our audience with today For consumers? Educate yourself.

Uh, if, if you are interacting with the bank website, if you are interacting with the e-commerce website, make sure that they have the right security controls. And if you don't see it, I guarantee that there are others who provide the right security controls, right? So to some degree, end users need to take responsibility themselves, but it comes with education.

And the second, um, important aspect is organizations need to take, uh, security very seriously. This is not just about generating revenue. Um, organizations have a responsibility to safeguard the end user's information.

And the best way they can do that is by having layered defenses in place. Um, and, uh, we do this at home, right? We have a door, a lock on the door windows, and uh, we have security cameras, et cetera.

When it's our own personal lives, we take all this extra measures. Uh, we should take the same measures when we are trying to safeguard the interests of, uh, consumers who trust us when they provide their information to us. So it's critical for organizations and enterprises to leverage the right tools, uh, to, one, protect their organizations, but more importantly protect the information of end users who are trusting them with that information.

So that's essentially what I would strongly advocate. Uh, there are many companies out there that, um, uh, are in the space trying to block or keep fraud stress out. Junior is one of them, but many companies are there.

And so it's critical that, uh, organizations leverage uh, one or a combination of the combination of these organizations to keep the process out and safeguard end users. Alright, well, thank you for coming on our show and sharing your insights with us today. Absolutely, Amanda, thank you for having me on.

And thank you to our audience. Stay tuned. There's more.