Onyx Security launched on Thursday with $40 million in funding from Conviction Partners and Cyberstarts, according to a company announcement, as the startup seeks to build what it calls a secure AI control plane for the growing use of autonomous AI agents in the enterprise.

The company said its platform is designed to discover AI agents, monitor their reasoning steps, and approve, block, or correct actions in real time, as businesses increasingly deploy AI-driven security and governance systems across functions including engineering and customer service.

Onyx said the rise of AI agents is creating new operational and security risks for companies, including reasoning errors, hallucinations, and attacks delivered through malicious prompts. The startup argues that traditional security tools were built for predictable software and human-led workflows, not systems that can act autonomously, a shift that echoes broader industry concern around AI-native cyber defense.

Chief executive and co-founder Maxim Bar Kogan said enterprises are already becoming “agent operators,” with AI systems gaining access to sensitive internal platforms and critical business processes. The company said its product is intended to give security, governance, and infrastructure teams a shared layer for oversight and policy enforcement.

Onyx was founded by Bar Kogan, a former Unit 8200 member, and Gil Elbaz, a veteran AI researcher who previously reported to Nvidia’s CTO and also served in an IDF AI research unit. The company said it already works with multiple Fortune 500 customers, employs more than 70 people across Israel, the United States, and Canada, and plans to use the funding to expand product and engineering teams.

The launch comes as Israeli cybersecurity companies continue to attract strong investor interest around AI-driven security and automation. Recent Jerusalem Post reporting has pointed to major funding rounds and record capital raising across the sector in 2025 and early 2026.

Cyberstarts partner Hila Zigman said AI systems are becoming part of organizations’ operational infrastructure, creating what she described as “an entirely new attack surface.” That framing reflects the broader industry view that agent adoption is moving faster than the controls meant to govern it.