AI, Agents, and the Future of Cyber Security

In just a few short years, the breakneck speed of advancements in AI have transformed nearly every industry, including cyber security. The pace of acceleration has forced IT and business leaders to rethink approaches to some of the most sensitive areas of their business operations, including workload management, innovation and DevOps, workplace mobility, and security.

Understanding how to plan and optimize for the continued evolution of AI and cloud deployments—as well as how to incorporate Agentic AI into cyber defenses and protect against malicious use—is now mission-critical for every modern business. Securing this new and shifting IT estate is a top priority not just for CIOs and CISOs, but the entire C-suite.

To begin helping cyber security leaders prepare for the evolution of cyber in the era of AI, Nadav Zafrir, CEO at Check Point, convened a panel of cyber security leaders at RSAC 2025, including Gee Rittenhouse, VP Enterprise Security Services at Amazon Web Services (AWS), Daniel Rohrer, VP Software product security at NVIDIA, Ami Luttwak, Co-founder and Chief Technology Officer at Wiz, and Nataly Kremer, Chief Product Officer at Check Point, to share their insights.

Agentic AI as the New Frontier in Cyber Protection

As AI capabilities advance, we are entering an era where machines can act autonomously, making real-time decisions without the need for human oversight. This evolution brings both promise and complexity. Agentic AI — AI systems capable of independent reasoning and action — has the potential to both protect and threaten the systems it operates within.

Rohrer emphasized just how transformative these new capabilities are: “The scale-out of agentic AI is incredible, but we need scalable trust models to align with these.”

He pointed out that while autonomous agents unlock immense power, we cannot afford to sacrifice visibility and governance in exchange for efficiency. “For fully autonomous agents, we will need more observability and boundaries. Human intervention or AI agent intervention will be needed to manage risk.”

Luttwak echoed this complexity, but also expressed optimism in the ways at which agentic AI could tip the balance in favor of defenders. “Agentic AI can give defenders more power because they can have hive knowledge. Once we see one attack, this information can be shared across the world instantaneously and these attacks can be automatically prevented.”

This “hive knowledge” model hints at a world where defenders collaborate at machine speed to stop emerging threats — before they spread.

Familiar Threats in an Evolving Landscape

Despite the radical advances in AI and security tools, adversaries are still relying on many of the same entry points they’ve used for years, including phishing and compromised credentials, giving defenders the benefit of continuing to train AI cyber security models to detect and prevent these threats, an area in which Check Point has shown a high degree of efficacy. While attackers continue to exploit these common vectors, defenders now have the opportunity to deploy AI agents to anticipate, detect, and neutralize them more effectively.

Luttwak added that the industry has the tools and knowledge to address this category of threats head-on: “There is a set of attacks in the world that are known, and AI holds the potential to fully eliminate these threats, which account for the vast majority of attacks.”

With AI’s capacity to automate detection and remediation, there’s growing hope that many legacy threats can finally be neutralized at scale.

Expanding Complexity Across Environments

AI innovation is not only changing how we defend systems — it’s changing where we need to defend them. With the rapid spread of cloud computing, hybrid architectures, edge devices, and traditional on-premises systems, the modern attack surface is more dispersed and diverse than ever.

Kremer addressed this reality head-on: “The scale of AI innovation is set to explode, and with it the cyber security landscape is going to expand exponentially. We have to protect cloud, hybrid cloud, on-prem, edge.”

Kremer also highlighted the critical differences in defense strategy across environments, noting:
“There are differences in protecting traditional versus AI workloads.”

Security teams will need to evolve their tooling and thinking to account for these distinctions, ensuring consistent protection across every digital layer.

Preparing for the Next Generation of Threats

One of the more sobering takeaways from the panel was the discussion around the speed of attacks in the AI era. When attackers and defenders both operate at machine scale, the pace of cyber conflict is no longer measured in days or hours — but in microseconds.

Rittenhouse challenged the audience to consider how real-time threats will shape the future: “How are we preparing for instantaneous attacks? We’ll probably have AI defend agents versus AI attack agents. In this scenario, things will happen at a different time scale than now. If this materializes, the entire cyber industry will change.”

He continued: “Additionally, patching strategies that exist today will no longer work because of the pace of AI. There will be a mixture where hardware will become more critical combined with agents sitting on top.”

Rohrer reinforced the need for infrastructure that can keep pace with these shifts. In his view, the industry must prioritize observability and proactive intervention — whether by humans or AI agents — to stay ahead of escalating threats.

Collaboration and Guardrails: Building a Safer AI Future

With the landscape changing so dramatically, our panelists agreed on one point with near unanimity: we need industry-wide standards and strong collaboration to ensure AI-driven security advances are implemented safely and effectively.

Rittenhouse emphasized the importance of collective action: “A joint standard for AI should be a requirement for us to move forward. I feel so positive about the collaboration and teamwork within our industry and among CISOs, and this gives me great confidence that we can create the right guardrails for AI.”

Kremer reinforced that security is a prerequisite for meaningful AI-driven innovation: “The biggest disappointment would be if we don’t use AI to dramatically improve productivity. But this won’t happen if it’s not safe — and cyber partners need to provide the tools to do this better.”

Final Thoughts: From Complexity to Capability

The convergence of AI, cloud, and cyber security marks a new era — one defined by rapid innovation, increased risk, and extraordinary upside. While the challenges ahead are significant, our conversation at RSAC 2025 highlights and reinforces a shared determination among industry leaders to adapt and lead with confidence into this emerging future.

As Luttwak summed up, the future of agentic AI offers reason for hope: “Once we can give the power and information about unknown attacks or zero-day attacks immediately following an attack, these attacks should then be able to be automatically prevented everywhere as AI communicates this information around the world. That’s my optimistic view for the future of agentic AI and security.”

As AI reshapes the foundations of cyber defense, businesses must evolve their strategies, adopt collaborative models, and implement secure, scalable frameworks. The time to rethink cyber security in the age of AI and cloud is not tomorrow – it’s now.

Watch the panel replay here.