"After taking the upper hand in AI security, we will use that to expand market share in the firewall market."

Lim Hyun-ho (임현호), who took the helm of security firm Checkpoint Korea last November, made clear he aims to turn the mood around in the domestic market with an AI-first security strategy.

He summed up the strategy as taking the lead early in the newly emerging AI security market and using that as a foothold to chase established players in the firewall market. To do that, Checkpoint has reshaped its network firewall product lineup around AI.

Lim said, "We are seeing results in our own way, and when we ran a proof-of-concept (POC) project last quarter, we also received feedback from customers that it was the most complete. It will be a winnable match," showing confidence.

Checkpoint rose to dominance in the global security market in the 1990s by commercialising firewalls, a representative network security product at the time. It later ceded leadership to Palo Alto Networks and Fortinet, which pushed high-performance firewalls based on dedicated hardware appliances.

When naming leading global security companies, Palo Alto Networks and CrowdStrike have come ahead of Checkpoint for quite some time, and the situation in Korea is not much different. That is the background to Lim using the phrase "turn the mood around" in his growth strategy.

To reverse the mood, Lim put AI security at the forefront as soon as he took over leadership of Checkpoint and delivered results by joining several large companies' AI transformation (AX) projects as a security partner.

He said the potential of an AI security-first strategy is large because Korean companies are emphasising speed in AX and many are addressing security alongside AX as part of risk management. Lim stressed again, "Checkpoint is now very different in character from the past. Our primary goal is to provide security services for AX that Korean companies are pursuing."

Checkpoint's AI security business is focused on 2 areas: Checkpoint AI Defense Plane, an end-to-end AI security platform, and an AI-based CTEM (Continuous Threat Exposure Management) solution.

AI Defense Plane covers both reducing risks that can arise when internal users use AI and helping companies strengthen security for AI-based services they provide to external customers.

For internal use, it provides management of AI outside administrators' control, so-called shadow AI, as well as AI red-team services. Lim said, "AI agents do not move on their own. They connect to LLMs and access external systems through MCP. Many security issues can arise in the process. Shadow AI is also becoming important from an administrator's perspective."

There are also many security considerations when companies provide AI-based services to external customers. Prompt injection, an attack type that inputs malicious instructions into large language models (LLMs) or generative AI to induce responses or actions different from the originally set rules or purpose, is already widespread.

Citing education as an example, Lim said, "Many educational AI tools are coming out, and if they start producing information that includes things like school violence, drugs or weapons, the problem can become bigger. If it gives irrelevant answers and a user is harmed, the service provider could also be responsible."

To prevent AI from doing what it should not do, companies across various fields are rushing to release guardrail security products. Checkpoint is no exception. It provides guardrail services through AI Defense Plane.

On this, Lim appeared to have much to say. He pointed out that guardrail security products placed in front of a model may block typical prompt injection attacks, but have limits in covering more advanced indirect prompt injection attacks.

Indirect prompt injection is a technique that hides malicious commands in external data that AI agents read, such as web pages, emails, documents and API responses, and makes the AI agent mistake them for legitimate instructions.

Lim said, "A method of putting guardrails in front of a model can block set phrases or specific content, but there are limits to blocking indirect prompt injection. You can see the phrases but do not really know the context. If you twist the context, you can gaslight AI." He added, "Checkpoint secured technology that can cover indirect prompt injection as well through its acquisition of Lakera."

In the case of CTEM, an exposure management (EM) solution takes the lead.

Checkpoint also provides, through EM, virtual patching technology that looks at traffic patterns before a vulnerability patch is released and blocks malware from operating.

Checkpoint in June also launched agent-based exposure validation for exposure management (AEV) to help defenders respond on equal footing against AI-based attackers.

According to the company, Checkpoint AEV uses AI agents that reason like attackers across an organisation's specific environment to correlate threat exposure data, asset context, real-time exploit validation results, threat intelligence and coverage, and determine whether exposed vulnerabilities can actually be exploited. AEV follows a safe validation loop instead of relying on static severity scores.

Checkpoint's strategy is to move into a more aggressive mode in targeting the market, judging that meaningful results are emerging in Korea's AI security market. It is particularly taking aim at the financial sector. With that in mind, it is also speeding up infrastructure investment for compliance with South Korean regulations.

Lim said, "Security in the financial sector has no choice but to go to cloud services. If you try to do it on-premises, the infrastructure cost itself is large, and there are also technical differences from the cloud. With the easing of network separation regulations, the financial sector will also be able to use cloud services as long as it maintains data residency, so we will respond actively."