Novee is an offensive-security company built around a proprietary AI system that performs penetration testing autonomously and continuously. Rather than relying on periodic, manual pentests or generic vulnerability scanners, Novee's AI reasons like a real attacker — mapping an environment, chaining exploits, uncovering business-logic flaws, validating findings with reproduction steps, and recommending precise, personalised fixes. Crucially, it re-tests automatically as infrastructure changes, closing the gap between point-in-time assessments and the reality of constantly evolving systems.
The company frames its mission around a sharpening threat: as adversaries increasingly use AI to attack at machine speed, defenders need AI that can think offensively at the same pace. Novee says its purpose-trained model was built specifically on offensive-security data and techniques, and reports that it outperformed leading general models such as Gemini 2.5 Pro and Claude 4 Sonnet by more than 55% on web exploitation tasks, achieving around 90% accuracy. This specialisation is the company's core differentiator versus general-purpose LLM wrappers.
Novee was founded by Ido Geffen (CEO), Gon Chalamish (CPO), and Omer Ninburg (CTO), national-level offensive-security leaders who came out of the Israel Defense Forces' most elite cyber programs. That pedigree, combined with rapid early customer traction, helped the company raise its round within roughly four months of inception — one of the fastest funding trajectories in the offensive-security sector.
The company emerged from stealth in January 2026 with $51.5 million in funding from YL Ventures, Canaan Partners, and Zeev Ventures. Its platform operates through a continuous loop — discover, detect, validate, fix, repeat — and the company lists customers including UiPath, JetBrains, K Health, Cresta, HiBob, and Telit, spanning multi-tenant SaaS and enterprise environments.
Novee enters a competitive autonomous-pentesting and offensive-security market, but bets that a model trained specifically for attacker reasoning — rather than adapted from general LLMs — will find more, and more meaningful, vulnerabilities while reducing false positives. Its trajectory reflects the broader 2026 surge in AI-native cybersecurity, where the contest between AI attackers and AI defenders is escalating rapidly.