Onyx Security

Onyx Security is a cybersecurity company that has developed a secure AI control plane to manage and govern the use of autonomous AI agents within an enterprise. The company officially launched in March 2026, securing $40 million in a funding round backed by investors Conviction and Cyberstarts. The firm was established by co-founders Maxim Bar Kogan, the CEO, and Gil Elbaz. Kogan has a background with the Israeli intelligence unit Unit 8200, while Elbaz previously worked as an AI researcher at Nvidia and in the Israeli army's AI unit.

Onyx's platform addresses the risks associated with the increasing adoption of AI agents, which can access critical systems and operate with a degree of unpredictability. The technology provides a unified platform that allows security, governance, and infrastructure teams to oversee AI agent activity. It works by continuously discovering AI agents, monitoring their reasoning processes, and intervening in real-time to approve, block, or correct their actions to ensure they comply with security and governance policies. This supervision helps protect AI systems from misuse and strengthens an enterprise's overall security posture.

The company's business model is centered on providing this single platform to serve multiple departments. Security teams gain visibility and runtime protection, governance teams can ensure compliance with regulatory standards, and infrastructure teams can manage and optimize AI agent deployments. The platform is powered by Onyx's own proprietary AI models and supervisory agents designed to understand AI reasoning. The company reports that it is already being used by Fortune 500 companies and will use its funding to expand its product and engineering teams to meet growing demand.

Keywords: AI control plane, agentic security, AI governance, autonomous AI agents, enterprise AI security, AI risk management, AI observability, runtime protection for AI, AI compliance, supervisory agents, Maxim Bar Kogan, Gil Elbaz, Cyberstarts, Conviction, AI agent monitoring, enterprise AI adoption, AI model security, AI safety, AI policy enforcement, security for large language models (LLMs)