Your employees are already using AI. The question is whether you can see it.
Across the mid-market, staff have quietly folded ChatGPT, Copilot, Gemini, and a dozen other tools into their daily work. Most did it without asking, because the tools are free, useful, and one browser tab away. For a lean IT team, that creates the same problem shadow IT created a decade ago, only faster and with higher stakes: usage you cannot see, on accounts you do not control, moving data you are responsible for protecting.
The good news for organizations already invested in Cato is that a real answer is available now. Cato’s new AI Security module, which came out of its acquisition of Aim Security, gives you visibility and control over how AI is being used inside your organization. And you do not need to wait for a full platform rollout to start. It can be deployed as a browser extension today.
Shadow AI is shadow IT with the volume turned up
The pattern is familiar. A tool solves a real problem, adoption spreads faster than policy, and IT finds out after the fact. What is different with AI is what gets exposed in the process.
When an employee pastes a contract, a patient record, or a block of source code into a free consumer AI account, that data can leave your control entirely. Free tiers often reserve the right to train on what users submit, and a personal account carries no business agreement, no data-residency guarantee, and no audit trail you can produce later. For an organization under HIPAA or PCI obligations, that is not a hypothetical risk. It is regulated data walking out the door through a browser tab, with no record that it happened.
The instinct is to block everything, but that rarely holds. AI genuinely helps people do their jobs, so a hard block pushes usage further underground, onto personal devices and personal accounts where you have even less visibility. The goal is not prohibition. It is governance: knowing which tools are in use, steering people toward sanctioned accounts, and drawing a clear line around what data can and cannot be shared.
Three layers, three jobs
It helps to understand that Cato addresses this across three distinct capabilities. They are related, but they are not the same thing, and knowing the difference is what lets you plan a sensible path.
CASB (Cloud Access Security Broker) gives you application-level visibility and control that a traditional firewall cannot. A firewall can tell you traffic reached a cloud service. CASB can tell you who logged in, what they did, and whether they uploaded or downloaded a file. It also lets you enforce policy at that level, such as requiring users to sign into ChatGPT with a company account rather than a personal one, or blocking unsanctioned generative AI platforms outright.
DLP (Data Loss Prevention) remains a valid, necessary layer of protection, and it focuses on the data itself. It flags and controls sensitive information as it moves across your SaaS and AI tools, so a file containing regulated content can be caught before it leaves. Cato has also extended DLP well past basic pattern matching: it can use machine learning trained on real documents and forms to recognize sensitive content by what it is, not just how it is shaped, and OCR to pull text out of screenshots and images. Older, string-matching DLP earned its reputation for false positives, flagging the shape of a credit card number whether or not the context warranted it. That reputation is fair for the older approach, but it is no longer the full picture. Learn more about Cato’s approach to DLP.
AI Security is the newest layer and the one built specifically for how people interact with AI. Rather than matching strings, it can classify the intent of a conversation, which cuts down the false positives that make older tools noisy and easy to ignore. It gives you visibility into the prompts people submit, the responses they get back, policy violations, and attempts to jailbreak or manipulate a model. It looks at both sides of the exchange, which matters more than it first appears.
Consider a tool like Microsoft Copilot connected to your SharePoint. Ask it the wrong question and it may surface sensitive information, such as compensation data, and hand it to a user who was never meant to see it. Watching only what employees type in misses half the picture. You have to watch what the AI sends back out.
Why you can start today
Here is the part worth underlining. CASB and DLP are advanced capabilities that run on the full Cato platform. AI Security does not have to wait for that.
AI Security deploys two ways, and which one fits depends on where you are today. If you are not yet on the full Cato platform, it runs as a standalone browser extension, working at the point where people actually use these tools, inside the browser. That means an organization can gain visibility into AI usage and interactions without first standing up sockets, rerouting traffic, or committing to a full platform migration. If you are already running the Cato Client and routing traffic through Cato Cloud, AI Security works in-network the same way your other Cato security modules do, no extension required. Either path gets you to visibility, and you can begin addressing AI governance as a task this week, rather than a project next quarter.
That distinction matters for lean IT teams especially. AI governance can feel like one more large initiative competing for time you do not have. Starting with a browser-based deployment reframes it as a manageable first step: turn on visibility, see what is actually happening, and make informed decisions from there.
A practical way to think about your AI strategy
If you are mapping out how to govern AI across your organization, a few questions cut to the heart of it. Which AI and LLM applications do you actually want to allow? Are your people signing in with company accounts, or with free personal versions that carry none of your protections? And do you know what kinds of data are moving into and out of these tools right now?
It also helps to widen the lens past the browser. Employees chatting with a web-based AI tool are only part of the picture. AI shows up just as often in desktop and command-line tools, coding assistants and IDE integrations among them, and those need coverage too. Cato AI Security reaches that ground through an agent deployed on the endpoint, or in-network via the Cato Client, so a desktop tool gets the same visibility as a browser tab. The same logic applies to internal applications that make API calls out to an LLM on your behalf. Those calls carry the same exposure as a person typing into a chat window, and Cato AI Security monitors them the same way.
Most organizations cannot answer the third question with confidence, and that is the real starting point. You cannot write a sensible policy for usage you cannot see. Visibility comes first. Once you can see the activity, the policy decisions become far easier to make and far easier to enforce.
The same thinking extends to any private or internal AI systems you run. Those deserve the same scrutiny as public tools, including monitoring for prompt-injection attempts and making sure internal systems are not sending or receiving data they should not.
Live Webinar | September 17
Watch Cato AI Security Run Live
Reading about AI visibility is one thing. Watching it is another. Spend an hour inside the Matrix production environment seeing exactly what this looks like: shadow AI discovery, sanctioned versus personal tenants, DLP guardrails on LLM traffic, and visibility into local agents and the tools they invoke.
Live from our own production environment, not a canned recording.
Jeremy Ness, CTO at Matrix Networks, with Liron Hakim Bobrov of Cato Networks.
No pricing pitch. A working demo and what we learned running it.
Thursday, September 17, 2026 at 10:00 AM PT. One hour.
Matrix Networks | 503-654-3000 | mtrx.com
Where Matrix fits
Matrix Networks has been working with the Cato AI Security capability as an early adopter, which means we are learning its real strengths and its rough edges firsthand rather than reading a datasheet. Our CTO, Jeremy Ness, has been leading that work internally so our team can guide clients through AI governance with practical experience, not theory.
For clients already on Cato, a fair first question is whether this capability is included in what you have or whether it is a separate addition. That depends on your current licensing, and it is exactly the kind of thing worth a short conversation rather than a guess. Reach out to your Matrix contact and we will walk you through what applies to your environment and what a sensible first step looks like.
AI is not slowing down, and neither is your team’s use of it. The organizations that handle this well will be the ones that chose to see clearly, early, and made simple, informed decisions from there.
Ready to see what AI usage actually looks like inside your organization? Talk to your Matrix Networks team about a Cato AI Security starting point built around your environment. We will help you turn on visibility first, then build the governance that fits. Visit www.mtrx.com or call 503-654-3000.
