AI Is Breaking Contact Center Security — Are You Ready?

AI is making cyber attacks easier to scale, more personalized, and harder to spot, and that’s showing up directly in the contact center. In this CX Today panel, Nicole Willing is joined by Miguel Fornes, Information Security Manager at Surfshark, Randy Layman, CTO at AVOXI, and Ron Zayas, CEO at Ironwall by Incogni. Together, they unpack how attackers are using agentic AI to accelerate social engineering, exploit the vast amount of personal data online, and increase account takeover attempts through higher volume and higher accuracy tactics.
The discussion also explores a fast-approaching shift: genuine customers using AI agents to contact support on their behalf, blurring the line between authorised and malicious automation. The panel argues that CX leaders need to rethink authentication as a risk-based continuum, not a binary pass or fail, with “graduated” security steps that increase as actions become more sensitive. The goal is to protect customers and frontline teams without turning every interaction into an obstacle course. 
Packed with vivid analogies, the panel makes complex AI-driven cyber risk feel real, practical, and urgent. Watch the full video now to catch every takeaway.

Transcript

SPEAKER_01 0:08

Welcome to CX Today. I'm Nicole Willing. Today's panel focuses on how AI is accelerating cyber attacks and what that means for customer journeys and frontline experience. We're seeing attackers adopt AI to scale and refine tactics like account takeover and social engineering. That shift is changing not only the threat landscape but also the cost, complexity, and emotional toll on customer-facing teams. So we'll explore how these threats are evolving, what it means for detection and response, and how enterprises can protect both customers and employees without adding friction to the experience. So joining me for the discussion are Miguel Faunas, who is information security manager at Surfshark, Randy Lehman, who is CTO at Oxy, and Ron Zayas, who is property expert and CEO of Ionwallby and Cogany. Welcome everyone. Thanks for joining.

SPEAKER_02 0:56

Thank you for having us.

SPEAKER_01 0:58

So ground the discussion, um, you know, when it comes to the customer journey, where are you seeing AI change attacker behavior most right now?

SPEAKER_02 1:08

I'll start off with um what our customers are seeing in their call centers today. It's it's really AI is making it so much easier, and the volume is just increasing. The tactics of most of the attackers are not changing very much. It's just so much easier to do it. Now, there are some sophisticated attackers that are trying more sophisticated attacks like deepfakes, but for the bulk of what we're seeing, it's just volume, more and more and more.

SPEAKER_03 1:36

And if I may add on to that, and and and Randy, I totally agree that the first thing AI is doing is volume. The second thing I think it's doing is trying to incorporate more information. And again, it's replicating what the human attack is. And the more I know about you, the better I can I can go in and do that. My biggest concern is the next generation of AI that'll be able to combine both the volume that you're already seeing along with the information that's out there, and then being able to make it in a realistic conversation, again, at volume. What anytime you can scale up like that, what you start doing is if your call center has 100 people, there's always gonna be 5% that are you know the most, you know, that are either not as well trained or you know are most anxious to help. That's always gonna be it. And you you're kind of hit or miss when you're coming in and doing five or ten calls. When you can do a hundred calls, not only are you going to be able to find those weaker links, but you're also gonna crowd out the ability for other people who have legitimate calls to be able to come in. So it's not just a scamming, it's taking away the bandwidth to be able to help your customers.

SPEAKER_00 2:51

And I would say that from my side, uh rather than uh focusing on on the customer journey, definitely customer journey is it's always one of the most important things for any decent company. But uh the fact that, well, customer always, most of the times is right, and it needs to be satisfied by the end of the day because after all, they are paying the money that supports absolutely everything. They are we are actually uh simply it's it's our livelihood. But what happens here is it's more interesting actually from the from the attacker's perspective. Now we spend uh decades, several decades, uh trying to uh educate the user on uh comfort, on convenience. I want to have things as fast as possible, as easy as possible, uh sometimes simply perhaps sitting on the couch. I want to do things that perhaps 15 years ago require a full-blown large desktop computer and a decent connection and being completely focused. Uh these this shift to to convenience uh is something that the attackers always try to hook on. And uh if we take this together, these uh people got, I don't want to say lazy, but we got used to the convenience of uh I want to do things in the least amount possible of clicks with a really good user experience and with a very beautiful uh with a very beautiful uh uh graphical interface. If we always want the easiest option, the the smoothest journey for the for the user, if we are insisting them that everything will be easy and uh everything will be rosy, perhaps we are lying sometimes. Because the truth is that with AI, uh AI, uh honestly, agentic AI AI is uh is a game change. Uh what before an attacker usually was spending uh like days or weeks to do to recover that all that data that everybody is simply uh throwing on the internet before it took a lot of time. Now with agentic AI that can be honestly done in the blink of an eye. So we moved from manual clunky attacks to uh to hyper automated shallow journeys. And uh well that the biggest change I would say that is efficiency. So that efficiency right now it's being uh it's being uh weaponized. Attackers are using the AI ruthlessly. It's it's it's always unfortunately the criminals, cybercriminals tend to be most of the time one step ahead. So I'm seeing the journey a little bit uh dark and shadow, I should say.

SPEAKER_03 5:55

Nicole, if I may add something on that to that too, just and I and I think when we're talking about the next generation of what's going to come in the future, I think the bigger concern, and Randy, I think you see this, you'll you'll end up seeing this long before we see all this. But what happens when the authentic consumer is also using an agent? So it's not just people using AI to attack you, it's people using AI to have to do the customer support for them, to be able to change it. Now you're gonna have to have customer-facing um people on your end being able to distinguish am I talking to a malicious agent, am I talking to a you know an authorized agent? Those are things that we need to be looking at today that are coming a lot faster than I think we anticipated.

SPEAKER_02 6:44

That is something that some of our customers are seeing today, um, especially in the travel industry, there are automated travel concierges that will make a call on your behalf to check the status of your travel. And our customers who are trying to employ synthetic call detection, right? They're looking for bots calling in. They can no longer treat a bot as malicious, right? They now have to consider who is that bot and what is the purpose of that bot. Which all brings us back to how do we know we're talking to the right person, right? That that is the challenge behind all of this. It's not just is it a machine or not, is no longer the deciding point.

SPEAKER_01 7:22

Yeah, that's so important. It's another point because um, as you mentioned, uh Ron, as well, you know, with AgenTech AI coming in, it's there's you know, the new ways for these malicious actors to infiltrate. Um, obviously, you know, social engineering is becoming one of the main methods because it's so effective. And now with AgenTic, and you know, this is already kind of a high impact area for the customer experience. So, how do you guys see AgenTech changing the way these attacks are executed, especially when it comes to social engineering?

SPEAKER_02 7:54

Yeah, I'll I'll take that one first. Um to me, I I see it as just more faster, right? Being able to have the agentic system, it's give it a target, give it the data that's available through social media, through the internet, through LinkedIn, through all of the things that are out there, these agents are now able to know everything about you. They can do the security questions. As we get deeper into some of the other exploits that are potentially coming, right? They can intercept your emails, intercept your MFA codes. And um, I'll echo Miguel a little bit. If if you take a certain look at this journey, it can be a very dark path that we're looking down. And I think that's what us in the industry need to be working towards is how do we help our customers understand a risk-based approach to authentication, to interacting with the customers, service the customers effectively, but also keep our own companies safe.

SPEAKER_03 8:49

And if I can lead on to it too, it's it's almost the opposite of what we've been doing in the past. So in the past, the idea was trying to minimize the human level. So, I mean, we we always have done social engineering. You know, I I worked with a CEO in the US many years ago. He was uh a very famous casino magnet, and he he had trouble seeing. So he had somebody with him, and when I would come to meet with him, for example, the person next to him would identify me and then give him all this background. You know, you you met with Ron last week, you talked about A, he talked about B, and when he came on, he was on it. You know, I mean that's social engineering. Um I went to the to the DMB, you know, to get my driver's license, and I didn't want to wait in the big line. So I kind of pretended like I was a very frazzled, you know, which I'm not. But um, and the person wanted to help. The person was being human, and they're like, no, no, no, no, don't worry about it. You know, come to this line and I'll help you from there. Ethically wrong, but what the heck can save me in our mind? You know, we and the idea for a lot of companies was how to take that human element. That's why we have security questions, that's why we have MFA, that's why we have all these other things. But what happens when the attacker now isn't human? We almost have to keep thinking in the other way. What can we ask that a human can answer or that a human would react with that a machine wouldn't know? And it's it's not only hard because somehow advanced agents have come in a very short time. Again, we haven't had as much time to adapt as I think we we thought we were going to have. But on top of that, there's so a lot of what we do is based upon information. Tell me your security questions, tell me what it is with your account, tell me what your last deposit was or whatever the case may be. And the truth is that AI has that information a lot more than you do. They have faster. I I don't remember what my last deposit was when my bank asks me that. An agent will know that. So, on top of us having to stop these uh attacks, we're gonna have to evolve to what is purely human, what is an agent, and again, what happens when a human uses an agent who's authorized? How do we know that we're giving credentials to the right machine? And the truth is that anything that we do to give credentials to a machine can be stolen or can be implied or inferred, and that's gonna be the goal. I I think that how we're gonna deal with this isn't even on the table today yet. I I think we're still trying to figure out what we're going to do, but I do think at the end of the day, how we control information is going to make the difference. If we choke, just like any LLM, if we choke them of information, we're choking their ability to do their job as effectively as they do.

SPEAKER_00 11:46

I I love the way that Ron presented it. I loved it because uh you were mentioning uh regarding the human, we were removing the human. The thing is that right now, what we created, what what certain companies created this advance, uh, it's so I would say groundbreaking that we are not anymore able to distinguish what is human and what not. So this is the the key point. And if for us engineers that we spend the last 20 years uh working in security, we have difficulties, very serious difficulties sometimes to to distinguish uh uh the actions of an agentic AI from uh the the the output of a normal of a normal user because everything is based on language, that's why they are called large language models. The thing is that it's absolutely incredibly difficult. It's incredibly difficult. Um and I loved also the way that Ron is presenting this regarding the information. You see, um again, I'm an engineer, I spend all my life on this, and and I'm also I'm also a Spaniard, and I'm also a man, and I tend to forget absolutely everything. Thanks God that my head is attached to my body and I'm not forgetting where I'm leaving the head. So what happens is that this more or less, everybody's like this. Who haven't forgot where the hell I left the keys of the car? Did I close the gas? Did I this that? We we forget. But what happens, and what on the point that Ron is absolutely right, is that uh agentic AI, AIs in general, LLMs, they don't work in that way. Humans, we work with something called episodic memory. We remember certain parts of things that passed that happened in the past. Me in particular, I forget everything. I'm I'm absolute I'm an absolute disaster. Unfortunately, uh an LLM uh doesn't forget nothing. Remembers absolutely every single thing that absolutely everything, everybody wrote on the internet. As long as it was publicly available, now it's on the it's on the uh on the training met on the training memory, on the top context of that LLM. Well that happens, yep, as Ron said, uh the LLM perhaps knows uh better who you are than you yourself because that third dog that you had when you were eight years old, with a little bit of luck you put that at a certain point on uh on a forum twenty years ago. If that forum still happens to be there, the LLM took that information. The LLM will push that over and with agentic AI, then we can say both the infinite knowledge that right now there was online with your own way or a perfect grammar and speech ability that it can perfectly emulate how a person from southern Manchester would be able to talk and send you a scan. Because to be honest, uh until now we were calling the fishing attack something called spray and bray. It was like volumetric, it was like statistic, it's imagine you are going, it's a massive giant, the massive giant net of the fishing boats, those trollers, that they are simply grabbing absolutely everything. What happens? They are not they were not uh targeted. Uh I remember like a few weeks ago I received an email from Warren Buffett himself. Funnily enough, uh written in German. Come on, that's ridiculous. That that is what the people actually is used to, uh to spot ridiculous cases of fishing. But nowadays, uh the same-cyber criminals, instead of launching that type of spray and prey attacks, what they are doing just simply is imagine uh automated style. Something it's uh it's people that uh they have knowledge, they have the capabilities, agentic AI, and they have all the absolute vast information that you during your life you left on internet, on LinkedIn, on forums, on Facebook, and it's available to whoever wanted to take. And I want to let you to say something which is for me, I always love to repeat. You know, decent companies have something called uh quality analysis. You know, it's it's one of the foundations, one of the current foundations of software development. You develop and then you have a quality analysis team that will start to check and bang against that to understand what happens. With uh AI and a genetic AI, uh, what I'm feeling personally is that the whole humankind was uh thrown uh into a case of beta testing of a technology that uh not very well, everybody understands how it works and the consequences of it. And uh yeah, unfortunately, we all of us we sign in to work as uh quality analysts uh without being paid. So, yes, it's an incredibly interesting experiment that we were all thrown into here just because companies say that all the available information that it was out there on the internet.

SPEAKER_02 17:29

I think that's one of the things that makes spray and pray so dangerous going forward, right? You have all that information, you marry that with AI, and you can still spray and pray, but you are spraying so much more effectively now because, like you said, I know the third pet you ever had. Um and that's a challenge, right? It's how how do you counteract that? Um I'll give an anecdote though. I think my father has come up with an amazing way to um counteract this, and that is he actually doesn't answer the question. He takes, he has a little algorithm that he does based upon what site he's on, what he uses it for, what what the question they provide is, and it's like it is something complete gibberish to answer the question. No one's guessing his security questions. He's managed to lock himself out a few times, but he's uh no one's gonna guess his security questions.

SPEAKER_03 18:22

And Randy, if I may, uh by the way, I tend to do the same thing. Now, the the downside, of course, is my father does the same thing, then my father unfortunately had a stroke and is non-communicated now. So for us to try to figure out what's going on is a whole different world. But I think you you bring a good thing of what we're gonna probably be seeing in the future. Number one is that what we give to a company and what we put out there may not be the same thing. We we may be intentionally lying. I don't remember what the term was, but when they used to make maps, people would have a street that didn't exist, and I forget what that was called. But the idea was that if somebody copied their map and put in the same street that didn't exist, that didn't exist, it had to be a copy. And that was it was the line that told you that somebody had done that. And I think that's the same kind of thing. The same way, because it's not just the large language model, it's not just the fact that they're able to suck up all this information, but remember, they're learning. So they they attack the company and they start changing, okay, I have all this information. Let me try type A, let me try to approach B, let me try approach C, let me try and approach D until I get it right and I start to learn from my successes. I think we're going to have to start interrupting that with lies, interrupting that with information above the idea of gibberish that wouldn't be available anywhere, along with understanding that the more you share online, the more ammunition you're giving to be used against you. And again, we haven't even started to look to see well, what do we do with the agents that we've empowered to do this, that we've given our credit card, our passwords, and everything else, and said, go on to do that. One last thing about this, and isn't so much using this to for you know, going after a company or going after Randy's customers or Miguel's frontline people, but what happens when somebody can influence the agents? So you may be training your agent to shop for you and do whatever, but what if they want them to only buy from United Airlines or from Lutanza and you think you're getting the best deal and you're not? Or what happens when you think you're giving your agent permission to find out about the best deal out there or get into your Netflix account and change it, and the agent is now being influenced to use that with your healthcare provider. At that point, the biggest challenge is going to be that companies, I don't think, are going to be able to do a lot about that. I think individuals are gonna have to take much more responsibility for that. And that's sad because they already have a lot on their plate with breaches and everything else that'll be going on and honestly.

SPEAKER_01 21:15

Yeah, that's such an important point, Ron, because you know, as we're saying, that there's you know, this line between what's a genuine and a malicious interaction is becoming blurred, and so it's putting more pressure on the customer, isn't it? Um so then what happens to the customer experience when the security risk increases?

SPEAKER_03 21:34

And I can I can jump on from our point, you know, we deal with very, very targeted individuals. So we're in our company, what we do is we protect the personal information of judges, of federal officers, of CEOs. So understand that when they call it an ass or something, we're paranoid to begin with. And and if we're not paranoid, they get paranoid because we're talking about their. Most personal and sensitive information, and we're supposed to be removing it. So, number one, we we kind of apologize up front and we say, I'm sorry you lost your password. This is going to be a difficult situation, but we put it into terms for them. We do this because if we made it easy and somebody gets this information, we're just giving them giving them the key to the kingdom. So there's an expectation, and then the person all of a sudden steps back and says, Oh, again. So I'm not in mind of Northstrom's. I'm not, you know, I'm not trying to get my Netflix account password back. They we want to have to, we we have to set different expectations for different levels of information. If I were a healthcare provider, I'd be doing the same thing, and I'm surprised them that more healthcare providers don't do that. Not just we want you to do this to be safe or we want you, but they need to say this is going to be an unpleasant experience because of how important it is to you. The other thing is that we try to tell them up front, and I think that's something that companies haven't been doing, and we're going to have to do in this new age, is before we send it up, not just say, pick a password and make it 72 characters. Don't just pick these questions, but explain to them what the factors are. So one of the things that we do is because we get a lot more interaction up front, is that we talk to the people beforehand and say, think about what happens when you forget your password. Think about what happens when you don't have this, and let's work backwards on how to protect you. What would be the best way? Okay, we're gonna use QFA. What if somebody else is stolen your email credentials? What if somebody can you know has put an agent on your phone to be able to cover everything that you're doing? What are gonna be the ways that we have this? And unfortunately, it's it's a lot more personal interaction, which is for us not a bad thing, but it is time consuming. But I think we're gonna have to look at the front end before people create, not so much automate that or at least put some intelligence to let people know what's going to happen if the worst happens. Work backwards rather than just pick your password, pick your MFA, move from there. I think we're gonna have to, the customer experience, as Miguel was talking about, is gonna have to start earlier than we normally started in anticipation of what happens when somebody tries to destroy your account.

SPEAKER_02 24:27

And Ron said something in there that I think is really important that we don't want to uh lose track of, and that is graduated experiences.

unknown 24:34

Right?

SPEAKER_02 24:34

So we're all accustomed right now to the call in, I authenticate, and then I can do everything. We're seeing more customers who are now uh accepting the fact that there is going to be some level of malicious intent and malicious action. And so they are now coming in and do a basic authentication, and you can get basic information. But when you want to do a next sensitive level, right, you want to issue a new credit card, you want to ship a credit card somewhere else, now let's actually take two or three more steps to make sure that you're really you. And I love the way I'm running talking to me of explaining to the user why. Um we've actually seen a couple of customers who said, you know, we have some level of request, so we don't actually care to give out the information, right? If I'm a hotel, I don't care if they're they're asking for adjoining rooms, that's fine. I mean, you know, we can say yes, we have them, without having to authenticate, right? So in a few scenarios, it's giving better user experience. And a lot of user experiences, I feel like we're gonna have to get to this graduated, you know, you start somewhere and then you have to elevate and elevate as you get more and more sensitive on different types of information.

SPEAKER_03 25:40

Randy, I think that that is great. And I you know, I use the banking experience for that. Banks put in as much security for you to make a deposit as for you to draw money out. And I always thought that was insane. I mean, I understand that people could use it maliciously, but money wanting to pay a bill, for example, is a much lower risk than me wanting to change the service address. Those two things are very different. Or to get a refund, you know, those are very different, and we're gonna have to get a lot better at breaking that path and not making a one united path with one individual that makes all the decisions for that.

SPEAKER_02 26:19

I think we need to really start looking at authentication not as a black and white yes and no, but as a continuum of probability. You are 95% likely to be Ron, or you're 25% likely to be Nicole, right? And that entitles you to different things. And the the yes and no is just so hard.

SPEAKER_00 26:40

That is that is quantum computing. It's it's the basis of the existing quantum computing theory. It's rather than one uh zero and uh and one uh transistor excited or or or or not without voltage. No, that is exactly the thing. And and the thing is that yeah, we are talking about customers and and and and they are always right and things like that. But the thing is that um uh what before panelists were discussing is for me has a name. It's education. And unfortunately, uh it's not going to get better. Uh what happens when with the customer experience? Well, that that that again that has a name on on software development. This is called friction. Whenever you are adding a lot of security, uh well, the usability goes down the hole because actually it's called like that. It's a security paradox. The most secure computer is the one that is switched off, buried under um on a bunker of 15 meters thick walls, not connected to the internet without nobody being able to touch it. That is secure. For why you then what's the point of having it? What's the purpose out of it? What is the usability? Zero, but it's really secure. So people, uh companies, they have we have an obligation of educating the user for the good and the bad. Otherwise, the user is, and I'm talking about everyone in general, we are going to suffer a lot of consequences and we will be hit and we will lose money. Uh our physical, our physical safety can be harmed. We need to teach the people. Because uh in here, I don't think that we are reinventing the wheel. We are not doing nothing with with AI. Even though it looks like groundbreaking, uh I would like to put it like very simple. Let's it in the end for me it's a tool. A tool like a car. What happens with the cars, you may say? Well, if we think back uh a few a lot of decades ago, perhaps by the end of the 19th century, anyone could have the anyone could drive a car. You could be drunk, your kid can drive, you can go and put whatever your your newborn in front of the of the steering wheel. There were no rules, there was nothing. Why? Because it was a very new technology and nobody actually was very understanding the consequences out of that. When more and more people started to jump into cars, this is where the friction started to increase, and the people recognized the risks. And people had to be educated, and governments, companies had to rethink how they are selling cars and how they are engineering those cars. Right now, absolutely nobody will think about uh the Noah Ford Model T. Doesn't have even those things, they didn't have even a security belt. I'm not even talking about uh driving assistants and stuff like that. But those things that we are right now giving for granted are product of uh decades of uh education, of uh paying attention what uh how the users are being harmed and the regulators as well, teaching and oblying companies to take certain uh uh certain steps to help the final user. Because in the end, generative AI, uh agentic AI, LLMs, they are again, they are a tool. Same as a car is a tool. So unfortunately, please continue, please continue.

SPEAKER_03 30:32

I was gonna say that the big and I love your analogy, by the way, is why I drive a 1970 MGB because it can't be hacked. And it's a British car, so it doesn't run, so it's very safe. But sorry to go. And uh, but you know, with automobile safety, we had decades to evolve. You know, as human beings, we evolved over, you know, you know, we we we evolved over a million years, and the things that we're used to are the things we're good. AI is asking us to evolve in days, it's asking us to evolve in months. And I think that when when we've been, when the comet hits and you don't want to be extinct, you know, you have to react very quickly. And I think that one of the things that is fairly new to human beings is how much information we have given out. I mean, we we've always been good about giving information because we we worked together. The only way to kill a lonely mammoth was for ten of us to get together and do that. And we had to share what our plan was. We had to share who was going to be in front, who was gonna run after it, who was the best at all this. So we had to share information. And social media companies and lots of other companies have taken that, taken that evolutionary trait that we have and just magnified it. I think one of the things that we keep coming back to here is that we're gonna have to fight that and evolve very quickly. And one of those is that you can't give information out the way you used to. You know, it took us a while to understand not to give out our social security numbers. I, you know, I went to school in the state of Nevada in the US. Your license had your social security number on it. And it took them a while to figure out that's not a good idea. You know, we we're training that to get people our credit cards. Um, you know, all of that, we have to start adapting that and telling people that any piece of information can be critically used against you. So that you have to be much better at pulling that information in because information is what humans use to validate each other. When I talk to Miguel, the first thing I ask him is, what's your position? What do you do? What's your background? Because that validates for me who he is and you know how he's on this panel. Same thing with Randy. We look on the LinkedIn to find the information of who that person is before we decide we're gonna talk to. Well, we're gonna have to bring some of that back in because that information is the key. Everything that we're talking about here is based upon knowing something that other people aren't supposed to know, but we're putting it out there. And I think that's gonna be the biggest change, Nicole, in what's going on. The company is evolving to not give out more information than they need to. Users have to retain that information. Sometimes they're gonna have to lie and specifically change that information, and hopefully we don't have to go back to the bunker of 15 foot deep walls as Miguel's been talking about, but you did describe my office at home. And um, and we need to make sure that we understand that our information is ours, and we need to control the information that we give out and use it for our purposes because if we don't, it's like giving out your social security number, it's like giving out a credit card number. You're inviting people, unfortunately, to maliciously use that against you. And I don't think we're there yet. I don't think, I don't think we've made, to use the car analogy, at some point we got to the point of saying we need to have seatbelt. You know, we want airbags, we don't want people to drink and drive. You know, we made decisions and understand that, okay, we want to give up a freedom in order to have something else. I don't think we're there yet with information, but we need to get there. Or otherwise, nothing that Randy and his team or Miguel and their teams do will fully charge, uh, you know, fully take care of all the issues that we have.

SPEAKER_01 34:29

I love all these analogies. Um, and the question that is kind of bringing to mind is, you know, there's kind of this concern in in the industry about, well, you know, we don't want friction, we don't want the customer, you know, to be upset when we're introducing extra security and questions. But um, I think, you know, Miguel, you refer to this. Do we just have to explain to them why this is so important and they're gonna have to adapt because it's you know, the impact of what AI can do is so significant.

SPEAKER_00 35:01

It's a must, honestly. Uh we are not discussing about um a minor vulnerability, a sort of vulnerability on uh Microsoft Office or whatever generalistic program that everybody is using on their on their most of the people is using on their computers. What we are talking is uh the foundations of um of how actually uh uh internet is it's being built, uh it's being shaped. And that sounds like incredibly dramatic, but but honestly, uh Paul Kruger, which was uh uh a finance uh a finance Nobel Prize, uh he one point one day said that uh the internet will have the the internet will have uh uh the same impact as the fax machine on uh international business. Obviously, it was at the very beginning. Uh I think that uh with uh with AI is exactly uh happening exactly the same. It's we are living in an in a bubble. We were forced into this, this is actually life-changing. Uh it was successfully being used to steal hundreds of millions of euros or dollars or whatever currency you want to put. It was successful, it's still right now increasing its success, and we are always several steps behind. So education is the key point. Because what Ron was saying before, uh I I I would put it like in a very old-fashioned sentence that everybody knows. Information is power. And nowadays it's more powerful than ever because the LLMs are trained on that information. That information was taken without asking anyone. It was they used, it was free to use, it was a buffet. You know? It was taken, nobody asked why. Very little companies actually they revolted. I remember Reddit made a little bit of a little bit of a boycott because they were seeing from where the wind was blowing, it was useless. They were not able to do anything, unfortunately. And all that information was was taken, was used to train the models, and those models, unfortunately, uh that technology is being used against us. So all that inform all the information that we gave away, all that power, right now it's in the hands of somebody else. Some super large companies, which it might or might not be fine. That's why we have regulators, that's why those companies are operating in jurisdictions that they are decent, and we have rights, and some others, some having more rights than others in Europe. We are really happy because we have things like GDPR. But um that technology is actually being used in with illicit proposals by cyber criminals. And the foundations are exactly the same. The Agentic AI works in the very same manner, using the very same context and using the very same information. That power, we gave it away. Right now, unfortunately, I feel that we are powerless. And to recover that power, aside the regulators, we need to educate the consumers about the risks of this. We need to we need to let them know of what happens with that information, and that we need to change, unfortunately, certain customs, and perhaps uh uh help the customers, the general population to uh include on their daily lives something that we call uh uh uh security in depth. These that sounds like really cryptic in security engineering is the concept that uh when you are defining your security architecture, when you are defining your enterprise architecture, you are always placing uh security controls, knowing and expecting that they will fail. Because if you have several liars, you know that the upper one will fail. Perhaps a second. So let's hope that the theory will resist the attack. That should that uh that that philosophy, that dogma that works so well for a lot of companies, I think that for general population, for the customers, should be actually uh embraced as well. We cannot trust anymore like the password, just in one password, as Randy was saying before. It's useless, it it's gone. It's honestly it's it's it's not working anymore. You need more security, you need more controls. Different uh removing your the information you have online definitely might be one. Uh ensuring that when you are browsing, you have any kind of security advantage like a VPN, something like this, that will help you at least to catch something, definitely indeed. Because again, what we are what we are discussing now is helping the the uh the average Joe, the regular population. And that actually I would say, well, so let's see what works for the companies. Companies are actually, decent companies are not actually being breached every now and then on a daily basis. It's because of a reason, and one of the reasons actually is it's this concept of security in depth. You need to put on your daily routines, on your daily life, perhaps you need to get used to use certain concepts like definitely multi-factor authentication, a good password, removing your information, removing the information that you forgot that you wrote 20 years ago, taking out, exercising your rights of forgotten, even depending on your jurisdiction, using a VPN, using an antivirus, ensuring that your devices are up to date, that that unfortunately it's an obligation of the of tech companies, and even I would say, even uh the govern governments, the education education bodies, they need to teach that. In the very same way that perhaps there we have sexual education on high school, perhaps it would be a very good idea to to give uh some uh some information security education basis on that, on that early stage, because we are helping actually, we are shaping up the the future generation to be less prone to be simply uh uh part of the attack.

SPEAKER_01 41:29

And then from the enterprise perspective, um, I was looking recently that Anthropic, um, you know, they've launched this uh project Glasswing because they've realized that their newest model, Mythos Preview. Um, you know, going back to what we've been saying about how these models have been trained on all this information with nobody's permission, and then they're saying, oops, actually this model is very good at infiltrating exploits that humans haven't even been able to detect for decades. Um, and you know, they're not they're saying they're not going to release this model because how of how potentially dangerous it is from a security perspective. So um, how should enterprise leaders be thinking about this threat?

SPEAKER_02 42:08

Well, it is a very, a very scary time, right? The the hype, I'm not sure how much is hype fact versus hyperbole, but some of the anecdotes that have been released are definitely concerning. But on the flip side, you look at what some of the largest hacks in history have been. I'm going to go back and not to pick on Equifax, but Equifax several years ago, they were running a system that had had known exploits for multiple years before it had been patched. And one of the biggest things is um, yes, like Miguel was saying, defense in depth is incredibly important, but also just keeping up to date and having good discipline about keeping your systems close, that today is um deflecting so many different attacks. As the cycle gets faster, we probably need to offensively use AI to be patching and to be deploying those patches at the same rate that others are using it in an attacking perspective.

SPEAKER_03 43:02

And I agree with Randy 100%. I think the you know, look at COVID. A new disease comes, spreads around, it shuts down governments, it shuts down countries, it shuts down businesses. And yes, I mean, we can't underestimate what developing uh a vaccine did, but what was the biggest tools we could use in the beginning? Washing our hands, you know, and wearing a mask. Um they're they're basic, everybody can do them, and we know forever. I mean, our moms have been telling us for generations wash your hands, you know, cover your nose when you sneeze, cover your mouth when you come. Those are the basics, and I think those continue to be. You know, don't talk to strangers. Um when we're on social media, we're broadcasting to strangers. You know what I mean? That's scary in itself. So I tell my child that somebody in the van and a puppy pulls up, you know, run because that's stranger danger, but feel. Comfortable to put everything on Facebook or everything on Instagram, you know, put everything on there. What the hell? What could possibly happen? We do this in our careers, we do everything else. I agree with Brandy that number one, the basics are there. Good hygiene is always good hygiene. When people ask you for information, ask why. I mean, when my children were small, every once in a while they ask a really weird question, and the first thing I would say is, why are you asking that question? And then they give me the history, and then I go, oh, okay, now I get it. You know, like, hey, what's your ATM pen, Dan? Why are you asking me that? Oh, because we learned about pens and oh, okay, got it. No, no, this is how things work. Um, so good hygiene, the basics are always good. I think the second thing we need to understand is exactly to Randy and Miguel's points before, friction needs to get added in. If I'm trying to buy from you, the friction should be minimal. We're just trying to make sure that the credit card is installed. We don't need to gather a lot from there. If I'm asking for more towels in my hotel room, I don't need to be authenticated to do that. There is no downside to that. If I'm asking you for another key, that's entirely different. If I'm asking you to re-key the key I have, because it doesn't work, could you please give me the key to room 22, you know, 30 and re-key it, that there's a lot more friction that has to happen. And then again, you tell the person what you're asking me for can be exploited in this way or this way. So we're gonna have to go through a lot to be able to get that. And by the way, that's also a matter of trust. One of the things that happens is that you hung, you know, just to pull some companies up, you know, pizza places, we'll ask you for pieces of information because we want to secure it, and then they turn around and they sell that information. Companies are also gonna have to understand that when you ask for something for security, there is an implicit promise that you're not gonna sell it. Because if you're using it for security and then you're selling it, you're making it available to other people who can use it for that. So companies are gonna have to understand that you can't monetize anything. And when I said that, I'm sure there are listeners in the US whose heads just exploded because the idea is to monetize everything. But if you want a secure world, there has to be trust that you say to a user, this piece of information that you're giving me will stay between you and us, and it's used for this reason. Not use it as a back door to be able to sell and get more from that. And then finally, again, users are gonna have to be very cognizant of the fact that giving out your phone number is just as dangerous as giving out your social security number today or giving out your credit card. If I know your phone number, I know where you live, I know where your family is, I start opening the door to all sorts of different information. So we have to educate. I love, we can't even get sexual training here, right, in high school, Miguel, so we'll work on it in the US. But giving some type of training that goes along with security, the same stranger danger that we taught kids, what's the stranger danger about your information and teaching adults that and understanding that with AI, they're gonna be moving so fast that we're gonna have to we're gonna have to adapt and use the basic tools because if we don't do that, I don't think we're ever gonna adapt quickly enough to be able to keep up with what's going on.

SPEAKER_01 47:24

No, exactly. Um I can I love I love this conversation, but in the interest of time, um so that we can wrap up, uh each of you could you um, you know, given all of these important points that have been raised, what's the single most important thing that leaders in the CX space should understand about uh you know the cyber risk of these AI developments uh in the context of customer experience?

SPEAKER_02 47:50

There are so many things. We've just barely scratched the surface in this, it's hard to pick out one, but I yeah, I I I think that the number one thing I would say is to start thinking about authentication not as a binary yes and no, but more as a continuum. And what is the action? What is the risk, and what's the negative downsides of the action, and how much authentication do you need to mitigate that risk? I think that's probably the one of the biggest things that people need to start thinking about that is not very common yet.

SPEAKER_03 48:22

And if I may then piggyback, I I love that authentication is not binary. I I love that. I think the second part in is don't use, don't collect more information than you use and respect the how you use the information that you get, that there are some things that you truly only want to authenticate, and you don't want to make that available because you're you're destroying the idea of you're using this to authenticate. And then the third thing is from the user's perspective, users have to understand that their information is personal and think about it. You can change your password. You know, you might be able under some circumstances to get a new password and get a new uh uh social security number. But when we start talking about biometric, when we start, if that is out there, remember to a computer is just numbers, it's not your fingerprint, it's it's a code that translates your fingerprints into zeros and ones. When that gets out there and it will, how do you change your finger? How do you change your eye? How do you change your face? So if you don't start protecting your information today, you're gonna put yourself into a situation or you're not gonna be able to do it.

SPEAKER_00 49:31

From my side, I would say that the most important advice I would say don't don't pretend, don't ignore this. Uh don't do the like the ostrich, you know, ostrich money over. Don't try to put your head, your head down the down the air, down the earth and think that you will be fine and safe. Um this AI changes the uh the IT and security business forever. It's not going to get backwards. Uh the thing of blast wing, project blast wing and mythos, it's actually a perfect case. It's incredible what what actually they are they are trying to publish. And the only thing I can say is that we are not uh we are trying to catch something like just like a ghost with uh chain link fence, uh traditional security is not working with uh with this. It's it's it's not going to work. Uh ramping up uh uh with a lot more you know human power uh uh security teams, it's not going to work as well. Because the fight has been absolutely, I would say that, reshuffled. And uh there's going to be an incredible digital divide between companies that are doing things agile, adopting AI, and those that are trying to ignore the case. The example that uh Randy uh presented before of Equifax, absolutely, it's literally, it's one of the perfect cases. It's straight away from the books of security engineering. Uh that's going to happen a lot, a lot more. And mark my words, you check the statistics of large breaches, including biometrics, as Ron is saying, which is absolutely catastrophic, those will exponentially increase. Why? Because uh with uh models like Mythos, uh the capabilities of cyber criminals to weaponize uh that inform uh that technology with all the information that is right now out there, uh they will always try to get to the lowest hanging fruit. So my advice don't do like an ostrich. Don't put your head below below the ground in with the case of AI, you are not going to escape, and try not to be the lowest uh hanging fruit on the tree. Do something. Embed on your CICD pipelines, on your processes, as much trustful AI as you can, because otherwise you cannot fight with humans uh this kind of ghost. We cannot trap this ghost with with our hands, with uh, with uh with a fence.

SPEAKER_01 52:11

Thank you all for all of your great analogies and uh essential advice for our viewers. I'm sure they'll they'll come away with some great takeaways. So thank you all for the discussion. Miguel, Brandy, and Ron. Thank you for having us on. Thanks so much. And to our audience for more interviews and articles on this important topic, visit CXToday.com, subscribe to our LinkedIn community, and continue the conversation in on our social channels. Thanks for watching, and we'll see you next time.