Opening Check Point Engage Paris, the security company’s CEO, Nadav Zafrir told 900 security leaders that frontier AI has collapsed the scarcity that once limited attackers – putting nation-state-grade offensive capability within reach of almost anyone – and leaving the industry “living between paradigms” with the old security playbook obsolete and no replacement yet in place.
Zafrir set the tone for the event with a caution to security leaders that the industry has entered a period without a settled playbook.
“We’re now living between paradigms,” Zafrir says. “The old map doesn’t represent reality anymore. The only way to navigate what comes next is to master the fundamentals while we build for a future we haven’t fully imagined yet.”
Zafrir described what he called the “collapse of scarcity” in cybersecurity – the erosion of the expertise, infrastructure, and capital that once limited who could mount a sophisticated attack and that is now increasingly accessible to anyone.
Jonathan Zanger, CTO at Check Point, went further, telling attendees the shift is structural rather than incremental: “AI has crossed a critical cybersecurity threshold. Vulnerability exploitation has gone autonomous, malware development has gone on-demand, and attack operations have gone industrial. That’s not a future risk, it’s the operating reality security teams are already living in.”
Zanger pointed to new Check Point Research findings to support the claim: the median time between a vulnerability’s public disclosure and its exploitation in the wild has fallen from roughly a year in 2021 to under a day in early 2026. And in a separate experiment modeled on UK AI Security Institute research, Check Point found frontier models capable of completing full, five-step attack chains autonomously, without human intervention at any stage.
Rebuilding network and AI security around autonomous agents
Turning to the company’s product strategy, Check Point executives outlined how both network and AI security are being rebuilt around autonomous agents. Chief product officer Nataly Kremer described a shift from static rule sets to AI agents that translate operator intent directly into policy in natural language, grounded in a prevention-first, scalable architecture that spans cloud, on-premises, and device deployments.
Avi Rembaum, president of Technical Sales at Check Point, brought these threads together into a single architecture for securing enterprise AI end to end. He argued that protecting AI in the enterprise cannot be treated as a single control point, but must span four layers: the perimeter, where AI traffic enters and leaves the organisation; the AI application itself, where models and agents are exposed to manipulation; segmentation, to contain how AI systems and agents move and interact inside the network; and the AI factory, where models are built, trained, and deployed.
Rather than a patchwork of point solutions, Rembaum made the case for a unified, prevention-first architecture that secures AI across all four, bringing AI, cybersecurity, and enterprise infrastructure together as one.
New research: The exposure gap is widening
The event also saw the launch of Under Pressure: The 2026 Exposure Gap Report, which found that critical vulnerability exposures more than doubled over the past year – rising to 42,6% of all critical exposures from 18,7% a year earlier – even as fewer than one in 12 vulnerability alerts proved urgent enough to require immediate action once properly validated against real-world exploitability.
Yochai Corem, VP and GM of Exposure Management at Check Point, says the gap stems as much from process as from volume: vulnerability management, security operations, and infrastructure teams too often work from separate tools and separate priorities rather than a single, shared view of risk.
“Security teams are flooded with intelligence, but still struggle to turn insight into action and reduce risk using their existing security investments,” Corem says. “Exposure Management closes that gap by combining real-world threat intelligence with safe, automated remediation helping organisations reduce risk faster while preparing for AI-driven attacks.”