Israeli AI cybersecurity startup Onyx Security has announced it has raised $40 million in seed and Series A financing rounds led by Conviction and Cyberstarts and with participation of angel investors from the worlds of cybersecurity and AI.
The company is developing a platform that aims to give enterprises control and oversight over AI agents, a rapidly growing field as organizations transition to using autonomous agents in operational tasks. Onyx was founded in 2024 by Maxim Bar Kogan, a veteran of Unit 8200 and former VP of Product and Engineering at Mixtiles, and Gil Elbaz, an AI entrepreneur who served in the Air Force’s Operational Technology Unit.
Onyx Security currently over 70 employees in Israel, the US and Canada, with its R&D center in Tel Aviv. According to the company, its product is already used by several Fortune 500 companies. The funding will be used to expand development and engineering activities and deepen integrations with AI and security systems in organizations.
The new layer of control in the era of AI agents
In recent years, organizations have begun to integrate AI agents into a range of operational activities, from software development to customer service and IT systems management. Unlike traditional software that operates according to predefined instructions, AI agents are able to make decisions in real time, learn from a changing context and operate various systems within the organization.
This capability opens up significant efficiency potential, but also creates a new layer of risk. An AI agent may gain access to sensitive information, communicate with customers or perform operations on critical systems. Traditional security systems, built for more static software, have difficulty tracking entities that operate dynamically and autonomously.
Onyx’s platform is designed to address this gap. The system continuously locates the AI assets in the organization, allows employees to define usage policies and implement them in real time. At the core of the platform is a mechanism called Onyx Guardian Agent that identifies abnormal activity by agents and can block dangerous actions, reduce permissions or require human approval before performing an action. “Organizations are becoming operators of AI agents today, whether they intended it or not,” cofounder and CEO Maxim Bar Kogan tells “Globes.” “Agents operate between different systems, receive permissions and perform actions at the speed of a machine. In such a world, point-to-point solutions or retrospective monitoring is no longer enough. Organizations need a central control layer that knows which agents are operating in the organization and what they are allowed to do.”
According to Kogan, when the company was founded, many in the industry believed that it would take time for organizations to adopt AI agents on a large scale. In fact, he says, the pace of adoption has accelerated faster than expected. “When we founded the company, a lot of people thought it was too early and that large organizations would only be adopting agents in a few years,” he says. “But in early to mid-2025, we saw a very sharp change. Suddenly, there was an influx of organizations integrating agents into coding, IT systems management, and internal processes.”
He continues, “In our data, we see an increase of between 10% and 25% in the number of agents being implemented in organizations every month. That’s a very dramatic pace.”
Bar Kogan notes that some organizations use agents developed by technology companies, while others build their own agents using friendly tools or in-house development. “In all cases, you end up with an autonomous entity that has access to a lot of important resources in the organization. Sometimes it can communicate with customers and sometimes it can perform actions in the outside world. Once that happens, you need both control and security.”
He says enterprises can already see a ratio of one AI agent for every ten employees, and in some cases even a higher ratio. “We estimate that in the future there will be more agents in organizations than people,” he says. According to Kogan, the Onyx platform focuses on three key dimensions. The first is security, meaning preventing dangerous or abnormal use of AI agents. The second is control over the policies of the agents’ activities within the organization. And the third is measuring and understanding the business value that the agents actually generate. “Organizations want to not only secure their agents but also understand where they are really creating value. The combination of security, control and measurement is what sets us apart in the market today,” he says. “We are entering a very exciting time in the world of technology and in human society,” adds Kogan. “People will work alongside AI agents, and we hope to be a critical part of the infrastructure that will allow organizations to do this securely.”
Hila Zigman, Partner at Cyberstarts said, “These are not just software tools to be protected – they are systems that make decisions, access sensitive information, and are integrated into critical business processes. From the very first moment, it was clear to us that Maxim and Gil understand the depth of this change and are building a solution designed for the AI era, rather than trying to adapt legacy tools to a new reality. We believe the Onyx team is exceptionally well-positioned to become a market leader in AI security for large enterprises.”
AI agents are quickly becoming the backbone of how companies operate. Employees are introducing increasingly autonomous AI agents to transform work across departments, from engineering to customer service. However, the unpredictable nature of agents creates a new operational risk for enterprises, as reasoning errors and hallucinations can lead to unintended actions. AI agent adoption also creates a fast-expanding attack surface that can be manipulated by attackers with specially crafted prompts. Traditional security solutions built for predictable applications and human-driven workflows simply cannot see or control these new “digital employees.”
To help companies face this new risk, Onyx is building the first Secure AI Control Plane to oversee autonomous AI agents and enable their safe and rapid adoption across the enterprise. The company’s unified platform continuously discovers AI agents, monitors each step of their reasoning, and approves or corrects agent actions ensuring enterprises are able to enforce security and governance policies.
Published by Globes, Israel business news – en.globes.co.il – on March 12, 2026.
© Copyright of Globes Publisher Itonut (1983) Ltd., 2026.
-1024x768.jpg "Red Snapper Used as Cudgel by Fed Judge")
