Zero Networks is making a smart point at the right time. AI governance needs to do more than show teams what’s happening. It needs to give them a way to control it.
With its new AI Segmentation offering, the company is going after a problem more security teams are starting to face. Seeing which AI tools employees use, or tracking where data flows, is one thing. Deciding what those tools are allowed to access, how they connect to internal systems, and where they need to be blocked, is something else entirely.
That gap matters because AI adoption is moving faster than internal controls. Employees are testing public AI tools. Companies are connecting automations through APIs. Security teams often only see part of the picture. And even when they do spot risky activity, they do not always have a clear way to contain it.
Zero Networks CEO and co-founder Benny Lakunishok summed up the shift in simple terms. He told ChannelE2E, "A lot of current tools can tell you which AI apps are being used, where data is flowing, or which prompts are risky. That’s useful, but it is still observational. Our focus is deterministic policy control at the network and access layer, meaning the ability to explicitly allow, deny, isolate, or constrain how users, devices, workloads, and data interact with AI services in real time."
Why this matters now
As AI agents and copilots become more common inside enterprises, they create access patterns that older security tools were never built to manage. When those systems move too freely across applications, workloads, and data sources, the risk goes up. That includes misuse, mistakes, and unintended exposure.
Zero Networks is pitching AI Segmentation as a way to bring least-privilege access to AI services, agents, and large language model environments. In practice, that means helping organizations answer a few simple questions. Which AI tools are approved? What are those tools allowed to do? Which systems should they be able to reach?
Lakunishok says this problem does not fit neatly into older security categories. He says, "This is different from traditional microsegmentation, which was built primarily around east-west traffic between servers and workloads inside the data center. It is also different from SaaS access brokers or browser controls that focus mainly on user-to-app access. And while zero-trust platforms define identity-based trust models, many stop short of granular runtime segmentation between AI tools, data sources, agents, workloads, and users. AI introduces new communication paths that most legacy controls were not designed for."
A bigger operations story
Zero Networks is also positioning this launch as part of a broader operational play. Along with segmentation and enforcement, the company is introducing an AI-powered compliance and risk engine. The idea is to let teams query network activity in natural language and map behavior to frameworks like NIS2 and CIS Benchmarks.
The larger pitch is straightforward. As AI becomes part of everyday business operations, security teams need tools that reduce friction instead of adding more of it. Still, buyers will judge this on execution, not positioning. The real question is whether these controls work cleanly in production and reduce risk without piling on even more policy overhead for already stretched teams.
Where partners come in
There is also a strong channel story behind this launch. Lakunishok said the opportunity comes from how fast customers are deploying copilots, LLM APIs, internal models, autonomous agents, and third-party AI tools. In many cases, adoption is happening faster than security teams can build consistent governance around it.
That opens the door for partners. Lakunishok pointed to services such as AI discovery, policy design, segmentation architecture, rollout, monitoring, compliance mapping, and managed enforcement. The business takeaway is clear. AI governance is starting to look less like a one-time policy task and more like an ongoing operational need.
He made the services angle even clearer. Customers are trying to answer basic but important questions. Which AI tools are approved? What data should they access? How should workloads be isolated? What will auditors want to see? Those questions create room for consulting work and recurring managed services.
That may be the bigger takeaway from this launch. As AI governance shifts from visibility to enforcement, partners have a more defined role in helping customers put guardrails around AI use in real environments. Lakunishok said MSSPs, MSPs, VARs, and resellers are central to the company’s go-to-market strategy. He also noted that many midmarket and enterprise customers will likely prefer to consume AI governance and segmentation as a service instead of building everything on their own.
For customers and partners, execution matters more than the message. Visibility, containment, and easier operations all sound good. The real test is whether tools like this make AI security simpler to manage in day-to-day environments, instead of adding one more layer of control to an already crowded stack.
