Horizon3's Adversarial Intelligence Frontier Lab: How AI Is Changing the Future of Cybersecurity

30 billion tokens a day. That’s how much real attack-and-defense data Horizon3 now collects from live production networks. Every command an attacker runs. Whether it worked. How the defensive tools reacted. How the services behind the firewall are set up. Today they’re putting that data to work. - Horizon3 just launched the Adversarial Intelligence Frontier Lab, and I sat down with Snehal Antani, CEO and Co-founder, at their San Francisco office to talk about it. - Snehal was CTO at Splunk and the first CTO of Joint Special Operations Command. When he talks about attackers, he has seen them up close. What stood out to me: - AI attackers are already here. They move at machine speed, and defenders need AI that can keep up. - The data is the edge. Most security AI learns from sensor telemetry, simulations or digital twins. This lab starts from attack paths that already worked on real systems, across 7,000+ customers from big banks to local hospitals. - Red and blue agents working together. One finds the path, the other closes it, and the hack, fix, verify loop gets faster. - Safety comes first. Agents running in production need scope limits, the ability to stop and roll back, and clear ways to measure what they did. - Critical infrastructure is a big focus. The lab will run Horizon3’s Cyber Resilience Centers of Excellence, where downtime has national and economic impact. The lab is led by Dr. Z. Berkay Celik, Associate Professor at Purdue and co-director of the Purdue Security Laboratory. 4 research areas to start: - AI-enabled attacks and adversarial behavior - Adaptive defense and attack-and-defense learning loops - Safe, controllable and measurable autonomy - Cyber resilience and operational readiness My take: whoever has the real data will build the best security AI. Horizon3 is betting that 7,000 live networks beat any simulation. I think they’re right. Watch the full conversation below and tell me: would you trust an AI agent to attack and fix your own network? Learn more about the lab: https://horizon3.ai/labs/ #cybersecurity #ai #agenticai #infosec #pentesting #horizon3 #securityresearch #theravitshow