Horizon3 launches lab for AI cyber defence research
Sun, 15th Nov 2026 (Today)
Horizon3 has launched the Adversarial Intelligence Frontier Lab, led by Dr. Z. Berkay Celik.
The lab will develop closed-loop systems that use evidence from production-safe NodeZero attack graphs to identify and close attack paths in live environments.
Dr. Celik joins as the lab's inaugural director while also serving as Associate Professor of Computer Science at Purdue University and Co-Director of the Purdue Security Labouratory. His work has focused on the security of complex software and systems, including vulnerability discovery, attack analysis, and the design and evaluation of defences for agentic and autonomous systems.
The research effort is intended to differ from many AI security projects that rely mainly on sensor telemetry or testing in synthetic environments. Instead, the lab will work from verified attack paths that have already succeeded against live systems, with a focus on keeping attack-and-defence systems safe, observable, and controllable in production.
"We've spent years collecting the attack-and-defence data required to research these systems. The next step is to use that verified attack-path evidence to improve how offensive and defensive agents work together in environments that actually matter," said Snehal Antani, Chief Executive Officer and Co-Founder of Horizon3.
Research agenda
The lab plans to begin with four research areas: AI-enabled attacks and adversarial behaviour; adaptive defence and attack-and-defence learning loops; safe and measurable autonomy; and cyber resilience and operational readiness.
In practical terms, this includes studying how frontier AI models may speed up vulnerability discovery and how emerging attacker tools operate, including coordinated agent swarms. The work will also examine how AI-driven attacks bypass existing security controls and how offensive and defensive agents can be coordinated through repeated cycles of hacking, fixing, and verification.
Another strand will focus on the design of autonomous security agents and the methods used to measure them. This will cover scope constraints, interruption and rollback mechanisms, and reproducible measures of effectiveness, defensive outcomes, and safety.
The final strand centres on critical infrastructure and operational readiness. It will examine how AI changes cyber operations and defence activity across operational domains, as well as the skills and training future cyber workers need.
Data base
A key element of the project is the volume of operational data available through the company's customer base. Horizon3 says it has more than 7,000 customers spanning banks, federal agencies, manufacturers, and hospitals, and that these organisations generate labelled training data from production-safe NodeZero tests.
According to the company, it collects 30 billion tokens of anonymised, aggregated, and labelled attack-defend training data each day. The dataset includes attacker commands, whether those commands succeeded or failed, how defensive tools responded, and contextual information on how key services are configured behind the firewall.
This gives the lab access to records drawn from live environments rather than primarily from telemetry feeds, synthetic settings, or digital twins. Horizon3 argues that this provides a stronger basis for studying real interactions between attackers and defenders.
The lab will also support the training of customer-specific cyber world models. These are intended to provide a shared context and long-term memory layer for red and blue agents so they can coordinate more effectively across repeated attack and defence cycles.
Critical systems
The lab will also oversee Cyber Resilience Centres of Excellence focused on critical infrastructure. These centres will apply the lab's research to essential systems in sectors where downtime or cyber compromise can have national or economic consequences.
This part of the effort is intended to translate research findings into operational improvements using production-validated attack-path intelligence and customer-specific models. The aim is to strengthen resilience in environments where service disruption has wider consequences.
Horizon3 has previously published research on AI attacker behaviour, cyber deception, and autonomous defence. That work has included studies on deception techniques for AI attackers, methods for evaluating autonomous cyber defence settings, and tool-mediated architectures for autonomous cyber defence.
The launch comes as security companies and researchers push further into the use of AI for both offensive and defensive cyber tasks. A central question for the sector is whether systems trained on real production evidence can be deployed in ways that remain observable and controllable while reducing exposure to genuine attack paths.