Introducing the Industry’s First AI Control Plane Designed to Oversee, Correct, and Protect Autonomous AI Agents
As enterprises move beyond simple generative search toward fully autonomous AI agents, what many are calling the Agentic Era, a new and unpredictable risk surface is emerging. Traditional security tools, built for predictable software and human-driven workflows, are struggling to keep pace with digital employees that can reason, act, and potentially hallucinate.
To address this, Onyx Security has officially launched with $40 million in funding backed by Conviction and Cyberstarts. Their mission: to build the secure control plane that allows enterprises to scale AI agents without sacrificing governance or safety.
AI agents are no longer just answering questions; they are integrated into engineering, customer service, and core operations. This autonomy brings two primary risks:
- Reasoning Failures: Hallucinations or logic errors can lead to unintended, destructive actions within a company’s production environment.
- Expanded Attack Surfaces: Attackers are moving beyond traditional exploits toward adversarial prompts designed to manipulate an agent’s reasoning.
“Every enterprise is becoming an agent operator, whether they planned to or not,” says Maxim Bar Kogan, Co-Founder and CEO of Onyx. “Agents are given access to critical systems, but we need guarantees they won’t make serious mistakes or get compromised.”
Onyx’s platform acts as a unified oversight layer. Powered by its own set of proprietary supervisory agents, the platform provides:
- Continuous Discovery: Automatically identifying every AI agent and Model Context Protocol (MCP) active within the organization.
- Reasoning Monitoring: Observing every step of an agent’s thought process in real time.
- Real-Time Intervention: Automatically blocking dangerous actions, requiring human approval for high-stakes decisions, or steering agents back toward safe operational parameters.
With a founding team hailing from the IDF’s Unit 8200 and Nvidia’s AI research division, Onyx bridges the gap between deep cybersecurity expertise and cutting-edge AI research. The company already counts several Fortune 500 companies as early adopters, signaling a massive enterprise appetite for tools that turn autonomous agents into governed participants.
As organizations scale from dozens to thousands of agents, the Control Plane will become the most critical piece of the AI stack, ensuring that the speed of innovation never outpaces the necessity of security.
