Elite AI Hackers Aren't Born. They're Trained.

The Novee founding team spent decades in offensive and defensive cybersecurity at the national level. They discovered and closed critical vulnerabilities that could compromise national infrastructure, and led operations that demanded the highest level of technical capability.

Through that work, they watched enterprises face an impossible bind: shipping code continuously while testing security only a few times a year, even as attackers operate around the clock.

In 2024, when OpenAI released the first AI that could genuinely reason through complex problems, co-founders Ido Geffen, Gon Chalamish, and Omer Ninburg saw that it was finally possible to build an AI agent that hacks with human-level sophistication. They founded Novee in May 2025 to clone their combined expertise into a proprietary AI system that runs continuously: finding the zero-days, business logic flaws, and chained attack paths that traditional tools miss, proving what’s actually exploitable, and closing the loop with verified fixes. That system trains and sharpens inside the Novee Gym, our continuous benchmarking and optimization environment. Since then, Novee has grown to serve dozens of enterprise and mid-market security teams globally.

The Novee Gym is where we test our full AI system – proprietary model, frontier models, and harness – across thousands of real applications, score every exploit, and use the results to continuously optimize for six metrics: precision (how many findings are real), recall (how many vulnerabilities we find), and quality (correct, reproducible, actionable) to be the best, and cost (low enough to run continuously), speed (hours, not days), and safety (no harm to your environment) to scale. As new frontier models ship, we evaluate them in the same Gym and route every task to the best-performing one.

The result is that you’re always ahead of attackers and AI, running the best model and the most specialized capabilities for every task.