NordLayer has a name for the mess of chatbots, coding assistants, and other AI tools piling up inside companies: an “AI zoo.”

Its new product, NordLayer AI Scout, is built to show employers just how crowded that zoo has become. The platform is designed to monitor AI activity across web and desktop, pulling approved and unauthorized tools into one view so IT teams can see what employees are actually using instead of relying on the software list they think they have.

“What companies are facing today is an AI zoo — dozens of unsanctioned LLMs in daily use across the workforce, each one widening the organization’s exposure to data leaks and security breaches,” said Andrius Buinovskis, VP of product strategy at NordLayer.

The phrase is colorful, but the behavior behind it is ordinary enough. An employee may use one chatbot because it writes better summaries, a developer may stick with the coding assistant they already know, someone else might open a free LLM because getting access to the company-approved tool takes too long. None of those choices has to feel like going around IT, yet together they can leave a company with a much bigger AI footprint than anyone formally approved.

That is the problem NordLayer AI Scout is meant to expose. NordLayer cites research showing that 70% of employees use free public LLMs at work, compared with 42% who use company-provided tools. A company can spend heavily on an approved platform and still have workers quietly building their own habits around something else.

View the NordLayer AI Scout demonstration here.

The company notes that NordLayer AI Scout monitors that activity on-device and measures actual use rather than assuming that a paid license tells the whole story. NordLayer says companies can see which tools have gained real traction, which licenses appear underused, and how much token spend is accumulating across deployed LLMs.

The visibility question gets more serious once employees start feeding company information into those tools.

A worker does not have to upload a document to create a problem. Copying a few paragraphs of internal material or pasting a block of proprietary code into a chatbot can put information outside the company’s control just as quickly, sometimes without the employee thinking of the action as a data transfer at all.

NordLayer AI Scout is meant to include data loss prevention built around the prompt itself. The solution can inspect what an employee is about to submit and apply company rules before the prompt goes through. It can warn the user or redact sensitive information, and companies can block a submission when needed.

“We built NordLayer AI Scout for how AI actually works — not just what it looks like from the outside, but for what’s happening at a technical level,” said Vakaris Noreika, VP of product research and development at NordLayer. “Its AI DLP is designed to inspect prompts and control them pre-submission, to ensure that sensitive data is never inputted in the first place.”

That gives employers another option besides simply banning unfamiliar AI tools. If workers are already experimenting, security teams can see where the activity is happening and decide which behavior actually needs a guardrail.

The picture gets stranger once AI agents enter the office. Coding agents can connect to outside resources and other systems through Model Context Protocol servers, meaning the tool an employee sees on screen may only be part of what is operating behind it. NordLayer AI Scout is designed to keep a live catalog of those MCP connections and inventory the skills available to each agent, giving IT teams a way to see what the software can reach after it has been set up.

NordLayer is betting that companies need that level of detail because the old model of approving one piece of software at a time is getting harder to maintain. New AI tools appear quickly, employees try them just as quickly, and agents can gain new connections after the initial approval is already done.

The company is not positioning NordLayer AI Scout as a way to keep every unapproved tool out. It is closer to an inventory of what has already moved in, paired with controls for the places where that activity starts carrying real risk.

For IT teams, that may be the more realistic assignment. The AI zoo is already open, now they have to figure out what is living in it.