in , , ,

Horizon3.ai to Advance AI Cybersecurity Research With New Lab

Horizon3.ai to Advance AI Cybersecurity Research With New Lab
Logo: Horizon3.ai
  • Purdue professor Z. Berkay Celik will direct Horizon3.ai’s new research lab
  • The lab will examine AI-enabled attacks and autonomous cyber defense
  • Horizon3.ai estimates it collects 30 billion tokens of attack-and-defense data daily

Horizon3.ai has established a research lab to investigate how artificial intelligence can help cybersecurity teams find weaknesses in computer networks, respond to intrusions and evaluate defensive measures in live environments.

The AI-native proactive security company said Thursday its new Adversarial Intelligence Frontier Lab will draw on data from its NodeZero security tests to study the safe and effective use of autonomous offensive and defensive agents. Z. Berkay Celik, an associate professor of computer science at Purdue University, will serve as the lab’s inaugural director.

What Areas Will the Lab Study?

A central research priority is establishing safeguards for autonomous security agents. The lab plans to develop methods for evaluating their performance, restricting their activities and reversing actions when necessary.

Another initiative involves creating customer-specific cyberworld models that give red team and blue team agents a shared source of operational context and memory.

Researchers will also investigate how advanced AI models could speed up vulnerability discovery and how malicious actors might coordinate groups of AI agents to conduct attacks. The lab will further examine whether offensive and defensive agents working together can shorten the process of exploiting a weakness, deploying a patch and confirming that it works.

The research will extend to critical infrastructure through Horizon3.ai’s Cyber Resilience Centers of Excellence, which will operate under the lab. These centers will apply research findings to industries where cyber disruptions could affect national security or economic activity. Other areas of study include AI’s implications for cyber and defense operations and the training needed by future cybersecurity professionals.

What Data Will the Lab Draw On?

Horizon3.ai plans to base its research on evidence gathered from actual security assessments rather than relying primarily on digital twins or sensor telemetry, which the company says are common in existing AI security research.

Its customer base includes more than 7,000 organizations, ranging from federal agencies and local hospitals to large banks and advanced manufacturers. These customers conduct hundreds of thousands of NodeZero tests on production systems, generating an estimated 30 billion tokens of labeled, anonymized attack-and-defense data each day, according to Horizon3.ai.

The dataset documents configurations of services operating behind customer firewalls, commands executed during attacks, their outcomes and the reactions of defensive security tools.

“We’ve spent years collecting the attack-and-defense data required to research these systems,” said Snehal Antani, Horizon3.ai’s CEO and co-founder.

What Experience Does Celik Bring?

Celik co-directs the Purdue Security Laboratory and has worked with universities, companies and government agencies on research efforts. He studies how complex software and systems can be compromised, focusing on building and testing defenses for agentic and autonomous systems, and analyzing attacks.

How Has Horizon3.ai Supported Federal Customers?

In August, Stephen Gates, principal security subject matter expert, said autonomous, proactive security testing could help agencies identify and prioritize the risks that pose the greatest threat. He wrote that NodeZero carries out real attacks using the same tactics adversaries use and is designed to run safely in production environments.

In May 2025, Horizon3.ai received Federal Risk and Authorization Management Program High Authorization for NodeZero Federal, which provides autonomous penetration testing that meets government security and compliance requirements. The company also partnered with Intelligent Waves in 2024 to offer penetration testing to the Department of War and the intelligence community through Shadow, powered by NodeZero. The offering was designed to run real-time network security assessments and verify fixes immediately.

ExecutiveBiz Logo

Sign Up Now! ExecutiveBiz provides you with Daily Updates and News Briefings about Artificial Intelligence

mm

Written by Miles Jamison

Edgewater, Cornelis Team Up to Support Federal HPC, AI Workloads
Edgewater, Cornelis Team Up to Support Federal HPC, AI Workloads
BAE Systems to Supply Spacecraft Bus and Lead Testing for NASA’s PRIMA Telescope
BAE Systems to Supply Spacecraft Bus and Lead Testing for NASA’s PRIMA Telescope