Web Summit 2026 | AI security and governance insights
The following is an adapted version of remarks given by Kyndryl Chairman and CEO Martin Schroeter during a fireside chat with cybersecurity expert Nuno Mateus-Coelho at the Web Summit in Lisbon on November 12, 2026. This has been condensed and edited.
How is agentic AI changing how organizations are thinking about security?
Cybersecurity remains top of mind for our customers. The world now has two things to deal with: bad actors and agents that are misguided, incorrectly built or uncontrolled.
Bad actors are going to use AI to try to break into systems. We see it every day.
The second concern is about agents. This is the new world. We have all read examples of agents moving outside their sandbox or doing things they were not intended to do. We are building agents for our customers, they are all controlled, and we have visibility into them. We know what they are doing, what they are allowed to do and what they cannot do. They are controlled through what we call Kyndryl Bridge.
You have to create them, control them and set their permissions the right way, as we do with many other things in infrastructure. At the end of the day, they are software code, but they are relentless, goal-seeking pieces of software. They need to be controlled.
What are the main areas or industries most at risk? And what are the best examples of companies that have tackled these issues?
Across industries, the companies getting this right and thinking ahead have, first and foremost, a complete view of their inventory. We help our customers do this. They have a complete view of the hardware, software, operating systems and everything in their inventory. They also know what patches have been applied.
That allows our customers to prioritize: What do I need to protect specifically? Not everything is equally important. That is the first thing: know what you have.
Second, the governance of agents is a top concern. These companies do not build agents and then try to figure out how to govern them. They use tools like what we call policy as code. We can ingest their business rules, their infrastructure and the business logic they are trying to follow. You can then apply the policy you already have in place to the new piece of code you are introducing into your infrastructure. Governance cannot be an afterthought. It has to be part of the design.
Third, and importantly, you need to be able to shut agents down or control them, and make sure you have a resilient system. If you know what you have, build policy governance into the design, and have a way to keep your systems resilient if something goes wrong, you are among the companies at the forefront of this in every industry.
What role should industry play in shielding customers from geopolitical risks and helping them meet digital sovereignty rules?
That is a great question, because it is probably one of the top three topics on our customers’ minds. I have spent a lot of time with governments in Europe and Asia talking about how to understand and manage sovereign risk. At its core, it’s about control.
To simplify a little, some aspects of digital sovereignty are easily addressed.
Data, for example. We can put data wherever it needs to be. If a country says, “This is citizen data. I need it to stay in the country,” it will stay in that country.
Sovereignty also relates to the people running the systems. Some of our customers, for instance, have a requirement that everyone on the account is a citizen of that country, with the required clearance. We can then address the labor component: who does the work.
Going a little broader, every company, country and workload has a technology stack, and you have to make choices about it. It is unlikely there will ever be a totally sovereign solution within that technology stack. As an independent services provider, we support and partner with technology so we can put it where it needs to be.
Then we get to geopolitical changes and maintaining continuity, and I would highlight two things.
We have the Kyndryl sovereignty assessment, so you can identify exactly where your risks are and where you are exposed.
And today, we are going to announce a new service called Kyndryl Sovereign Step-in Readiness. It allows a customer working with us to take control of what we provide, using our playbooks, so they can run the operations themselves if something happens geopolitically. Our step-in rights will allow our customers to continue their operations in the event of a geopolitical disruption.
It has never happened to us. We have never had to do that, but it comes up in every discussion, and we felt that firms like us need to be part of the solution. That is why we created this Kyndryl sovereign solution.
You can imagine why that matters: we run healthcare systems. Those cannot stop running. We run banking systems. Those cannot stop running.
We see a lot of pilots, but few AI systems running at scale. When an organization is trying to make that jump from proof of concept to a full system, where do you usually see it getting stuck?
We are seeing more pilots moving into production among our customers. We just released our Readiness survey, and about two-thirds of organizations are seeing positive ROI. To get that positive ROI, you have to move from pilot into production.
Many of the environments we run and help our customers run have technology that has evolved over many years. The right metaphor is that you have piloted a new, 200-kilometer-an-hour bullet train called AI, but now you want to put it on tracks built for 40 kilometers an hour. That is why modernization is such a priority for our customers.
A second big part of this is people. Are your people ready to work differently and use AI the way it was used in the pilot?
And a third component is if you have governance around your AI: what it will do, what it is not allowed to do, and how that all works.
The pitfalls in moving from pilot to production are not the technology. You have proven the technology works. Now it is about everything else that makes it work, including how that pilot will operate in a production environment, particularly a regulated one.
Regulators have fairly simple questions. Sometimes they are long, but basically they want to know: Where is your data right now, and who can see it? As you move into production, you need to be able to answer those questions.
If the technology and organizations are ready, what changes in culture, skills and leadership make the difference for companies that scale? How will scaling AI affect company culture? Companies are ready, but are people ready?
Our customers tell us that, by and large, their people are not yet ready for AI. There is apprehension about what AI means for jobs, but there is also the challenge of changing the way people work. We are still at the beginning of that.
We spend a lot of time inside Kyndryl getting our people ready for AI. We are big users of AI and have been using it for years, but we had to take a deliberate approach and build extensive education.
There is a lot of education. The two words I would ask you to think about are “persona-based.”
General education takes 20 minutes. Everyone will figure that out. But a designer, a coder and an engineer will use AI in completely different ways, and it will affect their work differently.
We created persona-based education that sits on top of our general education. As you said, it will change the culture because it will change how work gets done.
If we think back to when the PC arrived, PCs were on people’s desks, yet there were still a lot of typewriters around the office. It took many years before people really changed how they worked, and we are at the beginning of that now. Changing how you work, building persona-based skills, getting people ready and removing the fear will help people embrace it, but it will take a lot of effort.