Transcript
provider ecosystem, recording live at Mobile World Congress, Barcelona, 2026. From securing AI adoption to threat intelligence, business strategy, security operations, and value-added services, we bring expert insights to help service providers grow, stay resilient, and ahead of the curve. So let's dive in. I'm your host, Ronen Spirer, Director of Technical Solution Marketing at Fortinet, and I'm joined today by Cedric Jarowski of Transatel. Cedric, great to have you around. Thank you for being on the show. It's a pleasure to be here. Thank you very much. Thank you. Great. So let's start. We're going to talk a bit about AI and IoT and physical AI. Now, Transatel or Entity Transatel, you're part of the Entity group. You provide a large and wide range of services. What I want to focus is your global IoT connectivity services. Now we're seeing a lot of talk about physical AI, so the integration of AI in IoT in things. How do you see that happening in the real world? Do you see really that being taken up and used at scale? Oh, let's be clear. There is obviously a bit of hype in AI because everybody's talking about it, but it's not massively deployed yet. However, there is a contrast, if I may say. So we do see AI more and more used in the network itself, in the cloud itself, but on the devices, of course, there is a lot of solution that we can see around here at Mobile Congress on devices with Edge AI and capability to integrate AI in the device, but it's not yet massively available and deployed in the field. Because when we talk about IoT, there is a difference of product being on the shelf and product being actually on the field serving the purpose that they have to be. So on the network side, we do see and we implemented ourselves, let's say, AI-based anomaly detection being well used more and more by, let's say, mobile operator or service providers. And on the cloud side, obviously, this is where it all started. And AI is definitely also heavily used in order to provide some outcomes when we gather data of millions of devices. Right. So when we look at the future, so right now you said a lot of hype, very few things, if at all, at the field, Edge AI, physical AI. But let's just try and assume, and it will happen slowly, and it will happen maybe even faster than we think. Now today, there's huge quantities of devices out there, connected devices, they use specific standards, LTE-M, NB-IoT, but they're mostly, we're talking about huge quantity of devices, but relatively little amount of data. Do you see the possible integration of AI changing that? So suddenly we have much more data being used on these, and do you think that will change the requirement of the actual connectivity itself? Yes, indeed. I think this will have an impact. AI will definitely have an impact. But I would like to distinguish two types of use cases, if you allow me. So when we think massive IoT, so we will talk about trackers, let's say sensors, smart meters, this is, let's say, the big volumes of IoT. And this is where AI, back to my previous answer, is not always there because it doesn't make much sense, always, at least. And in this case, AI can actually play a role, but the impact of AI on connectivity itself for this category of device, I think, will be minimal. Why is that? Because, let's say, in those use cases, anyway, you transmit already a very small amount of data, right? So AI optimization, by providing algorithm at the sensor stage, can help, of course, can provide more intelligence, more efficiency, but it's not a major change, right? But where the needle might move is that for the most advanced use case, when we think about computer vision, robotics, let's say, on-the-edge decision-making. So for all of those critical use cases, yes, AI is already there. So this is why I have to nuance a bit my, let's say, my speech here. That's why we have you here. Exactly, yeah. And it so happens that for those use cases, yes, it will have an impact. Why is that? You have two factors to consider. The first one being that, obviously, through edge AI, you can optimize the amount of data that you can send because you can manage and treat some of those data locally on the device side itself, on the gateway, on whatever is on the field. So that has an impact. But on the other way around, it's also, let's say, AI can also generate sometimes more data and create more demanding requirements on the network because you have to exchange those data in a given performance. And those performance criteria can be latency, uplink throughput, you name it, right? And this is where, let's say, we might have to switch from the technology that you mentioned, which are more or less, they were born in the 4G era. This is where 5G is bringing some concrete elements in order to improve this concretely. How is that? But 5G is bringing the capabilities through slice, through a quality on demand to actually adapt those network characteristics to the constraint of the use case. So you can obviously derive higher throughput, lower latency through this type of technology. And obviously, let's say, AI can actually reinforce the requirement for those better performance. So this is why, for the use case I mentioned, like robotics, let's say, on the edge decision and so on and so forth. But I think what is even more important than those physical characteristics of the network and the connectivity is how AI actually drives the requirements to put more security on the network side. For me, that is a key point that is even more critical and more important than just the connectivity itself. Because with AI comes obviously tremendous opportunities, but also some risks that we have to consider. And this is where security embedded into the networks becoming even more critical. You actually went ahead of my next question, which was really about that. So how do you see, as you mentioned, how do you see, well, if you compare providing security services for AI, I'm sorry, for IoT today, compared to providing security services to AI-enabled IoT or edge AI, why do you see the difference? You said it becomes even more important. And what is your vision? Where do you think and how that should be provided? So the problem is that IoT is creating more constraints. Why is that? IoT devices have a tendency to be remote, to be harder to manage, harder to reach, harder to update. Also, and we know that in order to manage security threats or problems, we have to do some constant update of those devices. This is why they create an additional risk. So that's for IoT itself. But when we combine this with AI, and what is AI pretty much? It's the capability of a device to take decisions on its own and to trigger some events based on some data or some logic that are implemented on the device side. And this can create some operational risk up to, let's say, compromising entire systems. So this is why IoT combined with AI, definitely that's an opportunity to provide more value to the business, but it's also a greater risk that we have to mitigate with concrete solutions. And once again, our play here is to say the network is a critical part where you can embed security and provide the right level of service in order to overcome those difficulties of managing IoT device, but IoT devices combined with AI. So that brings me to the next question, sovereignty. So now when you say, well, you have IoT devices or edge devices, which are relatively stupid and simple and provide you a limited type of information compared to edge AI that can actually take action, it will provide you more intelligence and more information. Where does sovereignty get in the play? So sovereignty is a very good point, because most of the IoT projects, let's say, when the people in the enterprise design those projects, they miss some key elements. Of course, what we often educate them, please think about connectivity, security, and also data management in the first place, in your initial design, because most of the time, let's say when you design an IoT project, you start with the device itself, the sensors, how you will gather physically the data and so on and so forth. But those three factors I mentioned are always relegated at a later stage, which is not the best practice according to us. Because if you factor in those elements, yes, that might complexify a bit your initial design, but you will make sure that you will be covered for the lifetime of those devices. And we know IoT device can actually stay in the field for 10, 15 years sometimes. So it's absolutely the best practice to combine this. Back to your point on data sovereignty. So this is something that, let's say, we have been very careful about by deploying some IoT project in the world, because obviously, we have some specific relations that we have to face, especially in given industries. So in public transport, in, let's say, automotive, in energy, in not even speaking about public safety and defense, obviously. So in those sectors, we have regulations that we have to apply by deploying an IoT project. So the regulation might translate into constraints on the network side, but also on the data management side. So how do we cover this? Simply by having a distributed architecture that allows us to actually keep the data in the country where it's produced, which might be the key points to respect the data sovereignty rules. So in order to do that, we have built a distributed network across the globe in order to accompany our customers. And if you want your data to stay in South Korea, because this is obliged by law, they can stay there because we can issue the data over there, and they will be compliant with this. So it's in having, let's say, a specific network architecture that we actually comply with this. And obviously, we combine this specific architecture with what we call our security framework, which is articulated into three aspects. So the first aspect is protection. So protection is how do we combine our network and our connectivity capabilities with the right level of security. And for this, I already mentioned, but one of the key elements that we implemented that is working well is to have AI-based anomaly detection and threat management, because from the way the device is connecting to the network, from the traffic patterns, from, let's say, the resources trying to access, you can detect some anomalies and act on it even before it's creating the problems that I mentioned earlier. So the first was protection. The second was resilience. And to ensure resilience, this is the fact that we have a distributed architecture that allows not only to comply with data sovereignty, but also to make sure that we never have a single point of failure. And the third one is control. Control means that we master and we monitor our critical, let's say, architecture for that. And we do this because we have all our assets that we manage ourselves, core network components, connectivity management platform. And we expose some of those control capabilities that we do through our SOC and NOC to our customers so that they can co-control the solution with us and make sure that those AI devices that we're talking about will actually not, let's say, create more troubles than they will solve solutions. Cedric, it was fascinating. Thank you so much for your time and your insight and your expertise. It was great having you with us. Thank you very much. That's all for this episode of Fortinet OnAir, recording live at Mobile World Congress Barcelona 2026. Fortinet OnAir is available on YouTube, on all podcast platforms and Fortinet TV.
