The AI Extinction Event: Is Your Data Already Training Your Competitor’s Next Product?

Speaker A: Welcome to the Tech People podcast. My name is Ken Coyne. I'm your host and founder as well as an ambassador for OPS talent. I believe at the heart of any success story are the people who made it happen. Diversity, creativity and innovation. Where nurturing people can lead to an unbeatable formula. I created this podcast to share the experiences of some truly inspirational leaders on a journey to success. Enjoy the show. Hey guys, welcome back to the show. Imagine this. You just empowered your team with the latest AI tools to skyrocket productivity. They're smashing targets, moving faster than ever. While they're asking chat GDP to polish an email or summarize a meeting, they are quite literally copying and and pasting your company's crown jewels right out the front door. In highly regulated worlds like finance and law, this isn't just a minor slip up, it's a potential wipeout event for your business reputation. Today we're joined by a founder who argues we're living in the 1996 of AI security, a time before most of us even knew what a firewall was

Speaker B: or why we needed one.

Speaker A: Darren Way is the founder of Contextual and he's built the Alias Path, an AI data firewall that secures your sensitive data. We're going to on the Ghost machine or your most well meaning employees might be accidentally training your competitors next product and how you can stop the leak without killing the innovation. With that, let's welcome Darren. Hey Darren, welcome to the show.

Speaker B: Thanks very much King. Pleasure to be here.

Speaker A: No one likewise and great to connect recently. Well, like I normally do, it's always great to learn a bit more about you. I know that you've been working in data privacy since the 1980s. I'm sure you've got a lot of stories to tell, a lot of experience. Tech has changed quite a lot right since then and congratulations. I believe you're on your second business now. Maybe you could just talk a bit about your journey and where you've come from and where you're at today.

Speaker B: Yeah, absolutely, yeah. Um, as you say, I've been involved in data privacy and information security and all things technical really I guess since the late 80s. And yes, technology has changed um, a great deal but it's amazing how much stays the same and also the cycles that you start to recognize um, you know, as you're on that journey. But yes, second business now you know, previously had a business with uh, my business partner Rob West McCott. You're doing privacy orientated and elements but specifically for dealing with these and things like that. So helping organizations speed up their DSR processing, go through automation.

Speaker A: Okay, so tell us, I mean, how did you get into the whole data privacy area?

Speaker B: Well, you know, I guess it's one of those flukes of nature or whatever. You know, I was working in IT for a large insurance company, and I was asked to look at some aspects for a new system. And it had elements of your early, uh, stage data privacy, the 1984 Data Privacy Act. And I was looking at that, and I had to incorporate some of those things and pull some information from some mainframe systems as it was back then, to help this particular department meet their obligations. So grew to understand the act and understood what it was. And then after that, I became a little bit of a, you know, old. You know, speak to Darren. He's done a data privacy project. So when there was any new projects coming along, you know, people would come and say to me, oh, you know, what do I have to do? Uh, what's this data privacy thing all about? Uh, you know, because this was right at the early days of, well, only a few. Couple of years after the Data Privacy act had come into. Into being, that, you know, I was, um, I was talking to people about this. So and so through that, you know, that was that I moved on from that job. And then when I went into the next job, I had on my CV that I'd done some data privacy stuff. And you know how those things carry forward. You know, your boss comes to you and says, oh, Darren, you've done some data privacy stuff before. Can you help us with this or can you. Yeah, can you work with this department? They've got some things going on. So as is the life of, um, you know, of an employee, you, um, you know, you go to where you're needed and you do do the job that you need to do. And, you know, that combined with being highly technical and understanding information security and hacking and all those kinds of things just put me in a really good position to be able to have these conversations and to help different parts of the organizations. And at that time, I was working all in financial services. So highly regulated organizations like that are really on the front line with data privacy and information security, cybersecurity as they are today, but they always have been. Really.

Speaker A: Yeah. Okay. And, you know, congrats, by the way, on your first exit. I mean, it is. It's extremely difficult to get to an exit, and not many people do. I mean, was it very difficult to go again, Darren? Uh, and like, start again?

Speaker B: It really wasn't. I Mean, you know, it's difficult to praise Rob, uh, you know, but working with a good business partner, uh, is really what it's all about. You know each other's strengths, you know each other's weaknesses, and you know, through that being a good team together, you know, you fill in those gaps and you become, you know, it sounds trite to say it, but you become, you know, stronger than, you know, than the individuals can be. And so, you know, we work really well together. So once we finished, uh, you know, the exit, we had a handover period of a year that we worked with the organization that we sold to and once that was all done, you know, we had something very specific in uh, mind. So yeah, we decided that we were going to go again and you know, it was really a no brainer. I think we'd actually decided before the year was up that we were, you know, that we'd seen some opportunities and we were going to go again. So yeah, it was all um, you know, really good.

Speaker A: Okay, well done you. Well, I mean I'm sure over the years you've seen uh, some massive holes in terms of data privacy and data risk, but I suppose never more so has it been in the current environment because data is so important now, obviously with AI, LLMs and people sharing data, I mean what was it about the emergence of like the likes of chat GDP and these systems that made you realize this is the greatest, or would you, I suppose I'll ask you this, the greatest privacy risk of all time compared to previous experiences?

Speaker B: Yeah, I think it is. I mean, I guess that something that uh, CISO said to me, your Chief Information Security Officer, uh, um, I chat a lot with these kinds of um, people. Both um, you know, uh, because I've got friends who are in those kinds of roles, but also because they tend to be the people who um, we're talking to most about deploying your solution. But when I speak to those people, and one in particular about six months ago said to me, darren, this is uh, amazing, the amount of data that we're seeing go out of the building. I'd be looking to get some of these people fired if it was five years ago. But now the business is actively encouraging them to use ChatGPT. Claude, whichever LLM of choice the organization has, they're being encouraged to use it for a percentage of their role. You know, a percentage of their job now has to be done using their chosen LLM. And that's meaning that a lot of caution is being thrown to the wind. And this particular CISO was not quite pulling their hair out, but was, you know, very concerned about what was going out of the door and that they had no real control over. And the excitement and the glitter that comes with using, you know, AI and seeing the potential gains in productivity and things like that. It means that, uh, people and organizations are showing, throwing caution to the wind and really doing things that, uh, five years ago they wouldn't have, you know, they wouldn't have considered or wouldn't thought of. So, yeah, those are the kinds of things that we're seeing, you know, now. And that, in a nutshell, you know, sums up the concerns and the cybersecurity and information security risks, you know, that we identified and why little over a year ago, we created, you know, this product Alias Path.

Speaker A: Okay, well, there's a couple of points there I'd like to ask you about. Maybe just for the benefit of our audience, if you could simplify it. What is the privacy risk that you're talking about? Exactly. And then maybe we could talk about this new product that you built.

Speaker B: Yeah, sure. Absolutely. Yeah. So the privacy risks are really around every document that you send out to outside of the building, I very often say. But anything that goes out of the trust boundary. The trust boundary is sort of the highfaluting way of defining what is internal to your organization or where you consider a document to be safe. And anything beyond that border is, you know, the wild lands of the outside world. And once you send a document into the outside world, it's then beyond your control. Now, there may be contractual arrangements you have with an organization that means, oh, it's as safe as they can keep it. But as we see all the time, the organizations have breaches. You know, they lose data, they lose ip. In fact, you know, rather embarrassingly, Claude just earlier this week had a massive IP loss where they managed to release all the source code for one of their products, um, called code, wasn't it? Yeah, they released that out onto the Internet and, uh, made it available. So those kinds of breaches happen all the time. Even with the best will in the world, organizations can't keep your data safe. So our argument is, and our position is, you protect the data before it leaves the trust boundary and make sure it's safe and secure, even when it's outside of the trust boundary. And that's what the Product Alias path does, is it helps keep documents safe when they're outside, but still leaves them in a usable state, so they can still be used by large language models and can be used by AI they're not put beyond use, but their personal information in there is obscured. The technical name is pseudonymized, but it's obscured so that it can be, you know, the document can not roam freely, but can is uh, kept a lot safer than it would be if it was just released without any changes at all.

Speaker A: Okay. And you mentioned when we were offline about, you know, that we're in the 1996 of AI security where most people didn't even realize you need a firewall. Maybe for the benefit of the audience, would you just maybe expand the boldness and also then maybe uh, after that maybe going to, you know, why is the industries, you know, just by seem to be just bypassing all these checks and just leaving it happen the standard issue.

Speaker B: Yeah, of course, yeah. So I think as I mentioned earlier on the cycles or the products and businesses and all sorts of things going and obviously AI is one of the most quickly adopted products ever in human history. You know, the previous uh, products, um, you know that people talk about, you know, the telephone and the duration that it took to get a million users and then cell phones and the number of cell phones that uh, all the time it took for a number for number of cell phones in use to reach a million and things like that. We look at AI, it's been adopted far more quickly than even the Internet, which was previously, you know, one of those consumer products that was really taken up um, relatively quickly. So the comparison that I make in this is that I think we're round about probably about 1996 rather than 86, around about 1996. And the reason I think that is because around about that time organizations that I was working with, even those in the financial services sector, uh, many of them, you know, were adopting firewalls at that point in time. But they were still, they were still the kinds of um, firewalls that were rather rudimentary to our eyes now and, and many organizations still didn't have firewalls. Most organizations at that point in time were obviously still running their own email and things like that. And you know, it was not uncommon especially for smaller and mid sized businesses to have you know, a cupboard sized room with uh, you know, a bunch of, you know, desktop or tower PCs running most uh, of the, you know, the network and email and other services that that organization needed. And that feels a little bit like how things are at the moment. The one difference is that obviously the majority of AI is being used online. That's where all the frontier models are, um, the really big models of course, and I acknowledge and I know there'll be some people listening to this podcast saying, oh, I run AI locally and I've got this model running and things like that. And sure, there are a lot of people trying to do that. There are some organizations trying to do that. But the frontier models, in my opinion at least, are so much more powerful right at the moment. And only, um, growing. That gulf is to my eye is still, uh, very much growing. So technology wise and cybersecurity wise and data privacy wise was sort of about 1996 where some people are sort of saying, this Internet thing, is it safe? Ah, is it okay just to have a lot of, are all our documents out on the Internet? Is that okay? And you know, while some people are saying no, it's really not, we need firewalls and we need other forms of protection, there's still a lot of people are saying, oh, it's okay, just plug in, you know, just get online, just plug in, it's all okay. And you know, that's where I sort of see AI at the moment. Lots of people are still saying, oh, it's okay, let's just plug in some new processes. Oh, there's a new release that now makes everything agentic. Fantastic. Let's just plug in and go with it. Additional security, well, let's not worry about that. Let's just get the productivity gains and yeah, we can think about that afterwards. It sort of feels, it's a little bit laissez faire in that respect. Yeah, that's amazing.

Speaker A: Why do you think that is? I mean, is our, uh, company's under so much pressure to try and keep up with it or get ahead of each other or is just this ignorance or what?

Speaker B: I never like to think that organizations are, uh, you know, are ignorant. They're just taking. Their focus is just drawn in a particular direction and they don't understand the risks. And that's, uh, you know, that's a challenge for most organizations, right? There are so many things to juggle. You know, this is complicated. Your economies are in different states in different parts of the world. But, you know, essentially everyone's having to work that little bit harder to try and, you know, close the deals, etc. Etc. So anything that gives them a slight productivity age, and this flows obviously down to the individuals, because individuals are working that little bit harder. You know, they feel that they need to be, you know, putting in the extra effort. Uh, and if they can, you know, if they can do that through a shortcut by using chatgpt or Chord or whichever one they choose, then obviously they're minded to do that as well. And that introduces another danger of, um, what we call shadow AI as well. You know, where people bring their own AI to work, you know, would be on their phone, or they're bringing their own laptop into the office and things like that. They're, um. If the devices aren't locked down, then they can use their own, you know, versions of AI that they may or may not pay for. They may even be using the free versions, which is, um, you know, you know, doubly dangerous.

Speaker A: So in theory, they could literally take the phone, scan the document, or, you know, just copy the text into a private device, basically, and then run or choose the public LLMs, stuff like that.

Speaker B: We see examples of this when we talk to organizations. You know, you talk to the CISO or the CIO and they'll say, oh, the, oh, yeah, no, we don't allow LLMs within the organization. And then you'll talk to people within the business and they'll say, oh, yeah, that's right. It's all locked down. We can't use LLM, so I have to use it on my phone instead. So, you know, they will, they'll screenshot something, you know, take a photo of their screen with their phone, you know, push it into an LLM. Now that may, you know, the amount of data that may be transferred there may be comparatively less. But it's sort of that risk that you've got data flowing outside of the organization that again, a few years ago people would have been, you know, would have been fired for that. You know, perhaps they still are today if, uh, if the organization has banned the use of, uh, LLMs. But I haven't heard of any of those cases, I'll be honest. But perhaps some of your listeners have.

Speaker A: So I suppose in, in organizations where they feel, okay, let's just ban access to these LLMs, they think they have a solution. But in theory, people are obviously just finding ways around that which is like using their own private devices. And I guess it's very difficult to combat against that. Private devices.

Speaker B: Yeah, it really is. I think there's, I think there's probably three different types of organization. Well, three responses organizations are taking. The first is, and probably the most secure in some respects, but the most draconian is to say, right, no one can use these LLMs. We don't trust them, so you're not using them. So they get locked down. And there are many organizations that do that and take that approach. As we've Already discussed though, that leads to shadow AI, other organizations, and perhaps at the other end of the spectrum are the organizations who say, oh yeah, you're free to use your LLM so you know, it's the opposite end of scale. If you like, everything's completely open. They perhaps, um, one step, one notch back from that is, well, you can't use every LLM M, but you can use the one that we subscribe to. And that's probably where many organizations sit. You know, they might have Copilot or, you know, Claude, ah, or chatgpt again and you know, they'll subscribe to that and then they'll assign your users who have the. Have a business need will be given access to that. So those are the sort of, uh, those are three and I think there's probably a fourth, which is where we're trying to help organizations grow into those that are actually taking a, uh, you know, a proactive approach to AI security and actually ensuring that the data and the uses of that uh, of those AIs is according to the organization's policies. So actually keeping the organization that little bit safer, ensuring the data doesn't leave the trust boundary without their knowledge. Because even if you've got trusted users, you know, and you're giving them an account, you know, to use, there's still the opportunity that, you know, that their time pressure or a desire to get something, you know, something out the door, you know, before a long bank holiday weekend, for example. Yeah, it may result in someone, you know, sending uh, a document out to ChatGPT that in the cold light of day, they may have reconsidered the sense in doing that, you know, it might contain, obviously personal information, may, uh, contain corporate sensitive information. You know, it may contain, you know, any kind of, um, sensitive information for that matter. And without that being caught and pseudonymized and protected on its way out, all of that information flows out into the wild and you've got no idea then what happens to it next is one

Speaker A: of the biggest challenges for you, Darren, is it more on the education? Education like, uh, these organizations that they don't understand the risks with this or what is it?

Speaker B: Yeah, it actually varies because a lot of the CISOs, you know, that we speak to say, oh, thank goodness, I've been really worried about this, but we haven't, you know, we didn't have a solution. So there was nothing that I could say. And the business need is so great, but it's not something that I can stand in the way of. You know, we have a lot of those kinds of, a lot of those kinds of conversations. So with people like that, it's not so much education, it's just about, you know, helping them understand that we're here and, you know, what we can do to actually help them. But there are some organizations for sure that, you know, think that they're safe and secure just because the contract that they have with the AI provider, uh, you know, says that, oh, you know, the, the documents won't be used for retraining. And that's really only half of the problem because, I mean, there's an example that I talk about a lot which just is perhaps a little bit extreme, but it does sort of prove the point, but that are outside your control. So as many of your listeners may be aware, there's currently a court case in New, uh, York where OpenAI and Claude or Anthropic and another, another couple of companies as well. I think for those two key players are um, being sued by the New York Times because both those AI, uh providers ingested the contents of the New York Times newspapers. And what that has meant is that the judge has said that those organizations must retain searches and content to ensure that they can be reviewed by the New York Times lawyers for evidence that people are actually trying to read New, uh, York Times articles without actually purchasing. So by using, you know, chatgpt to actually read, regurgitate that information. So that means there's a, you know, a big store of information that's now being, um, kept and is not going to be used for the purpose that you as a user originally intended it to be used for. And that information is now being made available to third parties, you know, without your permission or without your say so. And that's just one of the examples. As I say, that's perhaps a little bit on the extreme side, but it's an example of how a document, once it leaves the trust boundary, is no longer in your control. You can no longer say, hold on a moment, no, that can't be included because it already is included and the judge has ruled it as such.

Speaker A: How about for those people in general that are just at home and working home and you know, they're doing, obviously doing everything in these chat GDPs and they're uploading whatever, just financial account, internal personal accounts, information, all types of personal documents. Is there anything those people can do, I mean, to address that or anything education that you'd say you'd advise these people?

Speaker B: I would. A long time ago. Again, uh, going back to, uh, the early days of the Internet. I remember having some conversations with some people who were, uh, brand new to email. And this may sound like a, an odd segue, but bear with me. These people had no idea about email. They'd never used it before at all. So they'd been given an account and initially they were just mailing each other because, um, you know, that was, they were the only addresses they knew. But obviously as more and more people, you know, in the country and in the world got email addresses and organizations began to swap email addresses like they did telephone numbers, then obviously people were then emailing things outside of the, you know, outside of the building, crossing the trust boundary. And that led to people, you know, saying things in email that they perhaps wouldn't have said in person or perhaps would have said in person. But obviously if they just said it, it was, um, you know, gone. It wasn't committed to paper and anything like that or committed to a document. Now my advice to people back then was treat email like a postcard. Theoretically it can be read as it traverses across, you know, back. But in those days, certainly it could be read as it traversed across the Internet. So treat it like a postcard. That doesn't mean that, uh, you know, you're going to be all jokey or, you know, isn't the weather lovely here? Kind of, uh, kind of messaging, keeping it serious. But treat it, this information may be read by others than the recipient, okay? And there may be multiple recipients and when that person leaves the organization, someone else may inherit their mailbox. Okay? So just think about those things. Really, the advice about using ChatGPT, any of the other LLMs is really pretty similar. You know, consider what you're sending and don't consider, you know, don't think that just because you're sending it to, you know, whichever MLM you choose, that that's its final and only destination. It may be being used in other ways. It may be being retained as in the case of, you know, the New York Times case. It may even be being used to retrain or fine tune as they, um, as the terminology goes, the AI so actually improving, you know, maybe that your content is being fed back into the system to improve the next version. Now you may think, oh, well, that's okay, there's no real worry in that. But if you're an organization with, um, commercial and strategic plans that are suddenly being ingested by an LLM, would you really be okay if with the right wording, someone could actually put in and say, I've got this plan to build this business that's very similar to your business. What are the best strategic plans or what are the most current, um, thinking around how to do this. And your information or part of your information is then revealed as part of that. Those are some of the concerns and some of the real areas of risk that exist in this new world.

Speaker A: Yeah, quite frightening, Darren, to be honest here. But get back to the whole what you built and the innovation aspect of it, just for the benefit of the listeners. Describe exactly how you guys are addressing this challenge.

Speaker B: Yeah, absolutely. I mean, put very simply, as you type in a prompt or add a file, as that information flows across the network, we take that information and we protect it through this process of anonymization. Let's just, um, call it anonymization. So that means, um, you know, and it's not just names and things like that, but I'll use names as the example because it's nice and simple for me and for others to understand as, uh, an example. So as the, the name goes out, you know, as Darren Ray goes out, then we change that name to be Fred Jones, for example. So the LLM only knows about Fred Jones when it receives a document and the prompt. So if I, you know, if I send a document with some information in, perhaps it's an HR record. And you'll be amazed, Ken, how many organizations are looking to send HR records out for. You're performing staff assessments and things like that. You wouldn't be on page.

Speaker A: Yes.

Speaker B: So that information is flowing out. So in this instance, you know, Darren Ray, and we change it to Fred Jones. So the LLM only knows about Fred Jones, knows nothing about Darren Ray. And perhaps my prompt says, oh, you know, what is Darren's role in the organization based on this document? Well, when chatgpt in this instance gets it, it says, what's Fred Jones role in the business? And all of the prompt and all of the document, any references to Darren Ray have been changed to Fred Jones and other types of information as well. You know, not just names. So that information then flows back through the LLM and it provides its answer back to us. It then passes through our, uh, reversal process and then Fred Jones gets turned back into Darren Ray. So as the user, I don't see any difference. You know, the LLM responds back to me. It gives me an answer, tells me what Darren's role is in the business. But the great news is that that information about Darren Ray and other information that was contained in the document has been protected as it's crossed the trust boundary to ensure that the LLM doesn't have that information and then can't draw upon Darren's knowledge or experience or whatever or at least can't associate that back to that individual in this particular protection use case.

Speaker A: Very cool, Darren. And is it, are you working only in certain industries or is it across industries? Where are the biggest risks you're seeing?

Speaker B: Okay, yeah, no we're across industries but those in the more regulated side of things, as with the firewalls back in 1996, those are the people who are coming, you know, who are actively looking for solutions that help them solve this problem. Yeah, because many of them are, are realizing now how much traffic is flowing out, how much information is flowing out, how many documents are flowing out on a daily basis. So, so it tends to be the more highly regulated organizations, you know, financial services, we speak to a lot of lawyers as well because you can imagine the kinds of information that are held in, you know, contracts, uh, you know, mergers and acquisitions deals, et cetera, et cetera that are flowing out on such a regular basis, you know, from law firms and financial services organizations. Uh, and it could be some really simple use cases as well. I mean uh, I'll just give you one, one example Ken, of an organization I was visiting with a uh, chief information security officer um, who was showing me around this financial services um, organization and this was in ah, a non English speaking country or where English wasn't the first language. And the staff were dealing with emails from you know, lots of different people from around the world. And obviously each individual didn't necessarily speak all of the languages that they were having to communicate back in. So they were using your chatgpt. I'm sure it was chatgpt, but wherever I say chatgpt, I mean, you know, or Claude or any of the others, but they were using ChatGPT in this instance I seem to remember, uh, to actually do translations for, for them. And even when they were replying back and they did speak the language they were having ChatGPT, you know, check it, make sure the grammar was correct and things like that. You might think, well that's a fantastic use for LLMs, and it really is. But the problems came where the contents of what they were sending was sometimes containing your bank account numbers, um, I even saw credit card numbers in there and things like that. And this information is all being sent, know, sent and received, um, you know, backwards and forwards through, you know, through an LLM and being parsed by an LLM. It's not the kind of information that most people Would want, you know, being, uh, sanity checked in that way, um, you know, irrespective of how they're then going to receive it by letter or email. However, there's then further protections that you know, should be obviously applied before you're using your system in that way.

Speaker A: Darren, one last question and I appreciate your time on this, you know, because I've come from a financial background myself. Financial services, working in banking and things like these. There's always a challenge in terms of, you know, innovation versus control and compliance. You know, you've got all these great ideas, you want to build new tech, you want to use the latest tech, but then you've got the bad guy, so to speak. Maybe guys like yourself to come in and say, no, no, we need to have compliance control here. We have to put this system in place. But does that system then slow down the whole thing? Does it, you know, Cause you got powerful AI tools you want to use and does systems and um, like other uh, controls or make it chunky, make it slow, ineffective. I don't know, how do you address that? Is that even a challenge?

Speaker B: Yeah, look, anytime I was considered to be convenience, uh, and security to be, you know, two opposite sides of a, you know, of an odd shaped coin, let's say, because, you know, really inconvenient to unlock your car, especially when you're carrying stuff, right? You know, it's really inconvenient. You're carrying a box and then you've got to try and rest the box against the window, you know, juggle for the remote control or you know, try and find the key and actually get it in the lock, et cetera, et cetera. Be far more convenient if we could just leave the doors open and leave them unlocked. But we don't do that. So anytime you add security to a system, you know, there is a small amount of change or inconvenience. Now in our system, you know, that's minimal. And we do everything to make that, you know, minimal and minimize it as much as possible, of course, but there is a slight change in that. But really the amount of time that an LLM takes to process information, it's not a sub second transaction that we're used to, especially in financial services. You know, when you're doing credit card processing or something like that, you know, you, we're very used to those kinds of transactions, you know, happening in milliseconds. But LLMs don't work like that. You know, they're thinking machine so, you know, they will go through a process. So the Amount of um, you know, the few milliseconds that we're adding to that process, you know, is not impacting the overall performance of the process. And certainly, you know, our customers and those who are, you know, trialing our product are very much saying that it's well worth the um, you know, the additional, you know, few seconds, milliseconds, um, to actually, you know, protect the documents on their way through the trust boundary.

Speaker A: Very cool. Sorry, one last question for you because you mentioned you're working across different industries. Are you only working with like large organizations or do you also work with small or medium sized companies also?

Speaker B: Yeah, certainly small, medium sized companies, we're working with those as well. It tends to be organized mostly medium and larger companies. But we would. Any organization who's got the, you know, got the need and recognizes the need within their organization, we'd love to have a conversation with them, of course.

Speaker A: Well, please feel free to uh, plug yourself, tell us how can people get in touch with you, your company, any other details, like to share, please go ahead.

Speaker B: Yeah, absolutely. So best and quickest, easiest way to find out more is to go to aliaspath.com perhaps Ken, you could include that in your um, in your show notes or in the uh, in the details for the podcast. But aliaspath.com is the easiest way to uh, find out more. And Darren at Contextual is my contextual IO is my email address. You can find that on the website aliaspath.com website.

Speaker A: Darren, listen, a pleasure to have you on today and it's very interesting because I talk when you told me about this, I read a lot about A.I. uh, I read a lot about security concerns, but all the security concerns seems to be around about the agents going rogue or protecting the internal infrastructure. And I mean to be honest, before speaking to you with the first person, really kind of highlighting the whole area of people actually feeding data into the system. So thank you for your time today in highlighting this week important topic.

Speaker B: It's been an absolute pleasure and thank you so much for uh, let me chat with you this afternoon and chat with your listeners.