As India’s cybersecurity market races toward a projected $3.3 billion valuation by 2025, fueled by cloud adoption and AI disruption, the role of forward-thinking security leaders has never been more critical. Sergej Epp, Chief Security Officer at Sysdig, brings a global perspective shaped by his engineering roots and deep engagement with emerging threat patterns. In this conversation with Tech Achieve Media, Epp shares insights on how CISOs must recalibrate their strategies to navigate AI’s dual role as both threat and tool, why open-source technologies like Falco have become indispensable in cloud-native defense, and how Sysdig is placing a bold bet on India’s talent and tech-forward future.
Also read: Sysdig Announces Real-Time Cloud Security SaaS Platform
TAM:What factors do you believe are driving the rapid growth of India’s cybersecurity market, and where do you see the biggest security gaps?
Sergej Epp: I believe India has tremendous potential in the new digital and AI-first world we’re building. One of the key reasons is the strong engineering mindset. Coming from a German background myself, I see how a culture focused on engineering fosters innovation and opens up many possibilities for the world.
If we look back over the past two decades, the entire software industry was made possible largely due to open source. A significant amount of commercial software was built on open source code. While India was part of that movement, it was not always among the earliest adopters. Still, a lot of innovation emerged from it.
Today, with AI and AI workloads becoming increasingly accessible, there is an even greater opportunity, and not just to build software, but to build entire companies. AI is powered by the cloud. You need to run these workloads somewhere, and you also need to manage and secure the underlying infrastructure. This is where we are now seeing strong traction as a company.
TAM: AI is increasingly seen as both an enabler and a threat in the cybersecurity landscape. How should CISOs recalibrate their strategies to responsibly harness AI while guarding against its misuse by adversaries?
Sergej Epp: When we look at today’s threat landscape, we see more than just AI emerging. Two major trends are accelerating at the same time. First, as you mentioned, AI can not only write code but it can also exploit code, which means it can scale attacks dramatically. Second, the nature of threat actors is evolving. We are no longer dealing with lone hackers operating from basements. Instead, we are facing well-funded, strategically driven nation-states targeting specific industries and even specific countries.
These two trends are converging, creating what many call a perfect storm in cybersecurity. One of the most common questions I hear is: who will move faster with AI, the attackers or the defenders? This is the new race. The challenge with AI is that its progress has been exponential over the past three years. The number of possible use cases is growing rapidly. Take, for example, AI-generated voice synthesis. While it brings benefits, it also enables a wave of fraud and new forms of social engineering.
We are now seeing attacks where someone might receive a call that sounds exactly like a loved one. Imagine receiving a call from someone who sounds like your mother, asking for help, only it is not her. That is a whole new level of manipulation. Then there is the sheer scale of attacks. AI boosts productivity for businesses, but it also boosts productivity for hackers. We are already seeing this play out.
Let me give you a data point. At OpenAI, every new model is tested and documented through something called a system card. When comparing models like ChatGPT-4.5 and earlier versions, the AI’s ability to solve “capture the flag” hacking exercises jumped from 20% to 80%, all within a few months. These exercises include things like writing malware or exploiting vulnerabilities.
Now, on some leading hacking platforms, the top performers are not even human, they are AI bots. So yes, attackers are gaining an advantage in the short term. But as defenders, we must ask: what can we do? The good news is that AI is also becoming part of the solution. We can use the same capabilities to enhance our defense, like supercharging our security operations centers. AI assistants can help us analyze massive amounts of data, identify attacks in real time, and even automate responses. The real problem in cybersecurity is not a lack of innovation. It is a lack of adoption. As an industry, we adopt too slowly. That has to change. Because in this AI-first world, speed is now the most critical factor in staying secure.
TAM: Open-source tools like Falco have become critical in detecting runtime threats in cloud-native environments. What are the key risks and rewards of relying on open-source for core security infrastructure?
Sergej Epp: I think Sysdig has been a pioneer in giving back to the world by releasing some well-known open source tools like Wireshark and Falco. For example, Falco now powers the cloud security infrastructure of more than 60 percent of Fortune 500 companies. It is also widely used by government agencies and even by some of our competitors in their own technologies.
As I mentioned earlier, open source plays a vital role in enabling global defense capabilities. But it is important to understand that if you rely on open source, you must also know how to run it. You need to take responsibility for keeping it updated, ensuring it is secure, and trusting the developers behind it.
Today, more than 80 percent of all code in commercial software is open source. So yes, open source is a powerful tool but only if you have the right technical capabilities to manage and maintain it properly. That is where we come in. Sysdig is, of course, a commercial company. We offer an end-to-end security platform for organizations that may not be able to manage open source tools like Falco on their own. Our platform simplifies the experience and extends its capabilities beyond detection and response to include vulnerability management, AI workload protection, and comprehensive cloud security.
Ultimately, it’s a strategic choice, which you must build and maintain it yourself, or buy a fully managed solution. Either way, that choice comes with responsibilities.
TAM: With the launch of your India SaaS instance, Sysdig has signaled a strong commitment to India’s data localization and cloud growth. How do you see India’s role evolving in your global innovation and go-to-market strategy?
Sergej Epp: India is a very strategic market for us. We were the first company to bring local cloud data center capabilities to India, well ahead of many of the large, established security players. From an infrastructure perspective, we invested early to be as close as possible to this emerging market. This allows us to offer our capabilities to a wide range of organizations, from startups to government entities.
Today, a significant portion of our APAC business already comes from India, and we expect to invest even more here compared to other regions. It is also worth noting that Sysdig, as a US-based company, generates a large part of its value and revenue from markets outside the US. We’ve earned global trust, in part because of our open source roots, and that has created a strong network effect across regions. India, in particular, stands out due to its deep engineering talent and leadership in software development. It’s one of the key contributors to our continued global growth.
TAM: From multi-cloud complexity to rising ransomware threats and regulatory compliance, what are the top three priorities that should be on every CISO’s agenda in 2025, and how prepared are organizations today to act on them?
Sergej Epp: If I had to break it down into three key priorities in today’s fast-moving, AI-first world, here is how I would frame it.
First, it is absolutely essential to understand your risk. This is where cybersecurity begins. But understanding risk does not just mean knowing which compliance frameworks apply to your organization. While compliance often drives security practices, what is becoming increasingly important is having the right mechanisms to validate your security. You need to determine which security controls actually matter for your environment.
This could mean running red team exercises or diving deeper into your cloud security setup to identify the vulnerabilities that are truly relevant. There is a lot of noise today, generated by tools, processes, and vendors, that may help with compliance checklists but do not actually improve real security outcomes. So it is critical to cut through that and focus on what really matters to your organization.
Second, once you understand your risk landscape, you need to orchestrate your entire security architecture around it. In cloud security, speed has effectively become the new zero trust. In the past, we tried to isolate networks and apply zero trust principles to reduce the blast radius. But in the cloud, which is more like a living organism, that static approach does not work. You need to be able to respond and adapt rapidly.
That means staying up to date with emerging threats and adjusting your security posture accordingly. For example, many companies today are adopting AI models like LLaMA, Mistral, or DeepSeek to drive business innovation, reduce costs, or disrupt their industries. But running AI workloads in the cloud comes with significant risks. These models are probabilistic, not deterministic, which makes them vulnerable to attacks like prompt injection.
The best way to protect against such threats is at the infrastructure level. These AI workloads run in containers in your cloud environment, and since they execute code and commands, they can be tricked into doing harmful things. If your containers are compromised, your entire system could be at risk. So security leaders need to adapt quickly to business needs using the capabilities they already have in place.
Third, identity management has become absolutely critical. In the cloud, a single identity or token can act as the key to your entire environment. In the past, attackers had to move laterally through multiple firewalls, an effort that could take weeks. Today, all it takes is stealing one key, and an attacker can access your systems from anywhere, often undetected.
At Sysdig, we have found that more than 97% of permissions in the cloud are not actually used. They have been granted, but no one uses them. At the same time, over 60% of containers powering applications live for less than one minute. The environment moves so quickly that it is essential to know who is doing what, when, and where. Identity is now the foundation for both controlling risk and enabling innovation.
Looking ahead, the next wave of innovation will be digital employees, AI agents with assigned tasks, running on AI-enabled devices. These agents will be powerful and efficient, but they will also operate aggressively and may unintentionally disrupt systems. That is why it is so important to assign them precise, limited permissions and clear identities. If we do not manage this well, we risk compromising our environments. At Sysdig, we see it as our responsibility to anticipate these shifts and help organizations prepare for what is next.
TAM: Final thoughts on innovations around explainable AI?
Sergej Epp: When it comes to AI today, I think there are two major trends worth highlighting. The first trend is the effort to create real boundaries, similar to how we once used firewalls. We now see the emergence of LLM firewalls that aim to control inputs and outputs, trying to prevent prompt injection attacks and similar risks. But mathematically, it has been shown that you cannot fully control the probabilistic nature of these models. They can always be manipulated or “lured.” So while these guardrails may help reduce risk, they cannot eliminate it entirely.
The second trend is the growing difficulty of securing the AI models themselves. Just look at Hugging Face, which has become a kind of GitHub for AI models. There are over 1,500 models available, and many of them powerful and freely accessible. But the problem is, you cannot really inspect what is inside these models. You do not know how they were developed or whether backdoors were embedded during training. With open source code, we have always had the ability to inspect, verify, and validate what we are using. But that level of transparency is currently not possible with third-party AI models. So you are left having to trust the model provider entirely.
From a security standpoint, this leads to one conclusion: you must assume breach. If you cannot fully trust the model or protect against prompt injections, you have to assume that the workloads you are running in your cloud are not entirely trustworthy. That makes it essential to reduce the blast radius, in other words, limit what these models can do if they are compromised.
Just as we’ve always focused on securing applications, we now need to secure the infrastructure, especially containers. This is where we, at Sysdig, come in. With capabilities like endpoint detection and response, container security, and API protection, we can help reduce the risk posed by AI models and agents. It’s critical to limit what these models can do through APIs as well.
Technologies like MCP (Model Control Plane) and other emerging innovations are exciting. They bring powerful new capabilities, but also introduce a whole new class of threats. It is a double-edged sword. At Sysdig, we are actively working on our own MCP server and AI agents. In fact, we released our AI assistant, Sage, two years ago. It is evolving steadily. Right now, Sage helps provide contextual awareness across cloud and Kubernetes environments. It starts with analysis, but we can already see its potential to perform autonomous actions in the near future.
But before we enable that capability, because technically it is just one click away, we need to ensure that it won’t “break things.” That trust comes from two key elements.
First, we need the right data. This is where we believe Sysdig has a unique global advantage. Data builds trust, and trust enables safe automation. Second, any autonomous actions must be secure. And that is our responsibility as a security vendor, to ensure that AI does not just function well, but functions safely. Ultimately, AI for security and security for AI will be two foundational pillars driving the future of Sysdig. They will shape how we evolve our platform and grow our ecosystem, alongside our open source community.
