Horizon3 Launches Adversarial Intelligence Frontier Lab

The lab, led by Purdue's Z. Berkay Celik, draws on data from Horizon3's NodeZero penetration testing service used by more than 7,000 organizations.

Updated on Nov. 12, 2026 in Cybersecurity

Horizon3 Launches Adversarial Intelligence Frontier Lab

Horizon3 AI Inc. launched its Adversarial Intelligence Frontier Lab, with Purdue University's Z. Berkay Celik named its inaugural director. The lab draws on data from NodeZero, Horizon3's penetration testing service used by more than 7,000 organizations.

Why it matters

The lab is expected to speed up development of Horizon3's red and blue security agents, a step toward autonomous testing of live networks. Horizon3 said its May paper found its AI system never proposed an action outside its approved tool list across more than 420 deployments.

NodeZero tests generate about 30 billion tokens of anonymized data a day from customers' live production networks, and the company has run more than 350,000 production-safe tests to date. These are Horizon3's own figures, and the extraction does not say how the lab will validate them.

The players

Horizon3 AI Inc.

A cybersecurity company founded in 2019 that sells the NodeZero autonomous penetration testing service.

Purdue University

A public research university in West Lafayette, Indiana, whose computer science faculty host the new lab's director.

Z. Berkay Celik

An associate professor of computer science at Purdue University who co-directs the Purdue Security Laboratory.

The details

Horizon3's cyber world models are trained separately for each customer and give red and blue agents a shared memory. The May design limited the AI to a fixed list of approved tools, using Anthropic PBC's Claude Sonnet 4 and Claude Haiku 4.5 models.

Timeline

  1. May: Horizon3 published a paper on keeping autonomous security agents under control.

  2. August: Horizon3 raised $250 million at a valuation of more than $2 billion.

  3. October: Horizon3 said it passed $100 million in annual recurring revenue.

  4. November 12: Horizon3 launched the Adversarial Intelligence Frontier Lab.

The Tech Race

The lab extends Horizon3's NodeZero autonomous penetration testing service by turning its live-network test data into training material for red and blue agents. It places the company's autonomous security work under university research oversight as the sector moves toward AI-driven defense.

The lab's work is aimed at making penetration testing of live networks more automated, which could change how often organizations test their systems. The extraction does not say when any resulting tools will reach customers or what they will cost.

The takeaway

Autonomous security testing is moving from research toward live production networks, and organizations that use such tools should watch how labs like this one validate their safety claims. Horizon3's reported safeguards are a useful benchmark for judging future products.

Further reading

Explore more Cybersecurity coverage for context on how AI is changing network defense.