AI Exposure Check · free for companies with 50–500 people

Your people are already using AI. Do you know what it can reach?

Staff paste company data into chatbots, connect AI note-takers to mailboxes and build their own agents. The free check shows your AI, what it can reach, and the fix.

Get the free check

67% of people using AI on work devices used accounts that weren’t the company’s. Verizon 2026 Data Breach Investigations Report

Request your free AI Exposure Check

About you

About how many people sign in at your company?

Rather talk first? Book a 30-minute call.

  • Every AI tool, AI app, agent and MCP connection, in one list
  • Who uses or owns each one, and which have no owner
  • What each can reach, ranked by risk with the fix

Free and read-only. Built on Centeye Lumen (AI governance and registry) and Centeye Synthwatch (early access).

The problem

AI arrived faster than anyone could set rules for it.

People use AI to save time, and most of it happens out of IT’s sight. Here is what that looks like.

67%

of people using AI services on work devices were using accounts that weren’t the company’s, per Verizon’s 2026 Data Breach Investigations Report.

What goes into those accounts leaves your control.
1 in 3

accounts hold an admin role at a 9-person business we watch. The smallest teams have the most admins for their size.

An AI app or agent running on one of those accounts can reach everything that account can.
Automatic

Microsoft Learn says Copilot Studio creates a Microsoft Entra Agent ID for each new agent automatically.

Agents appear as soon as anyone starts building, whether IT planned for them or not.

Verizon 2026 Data Breach Investigations Report. Admin figure: Source: Centeye platform data across the businesses Centeye protects, October 2026. Sign-in attempts counted over the last 30 days. No client is named. Copilot Studio detail from Microsoft Learn, November 2026.

Centeye Lumen Discovery screen: AI in use, people using AI, high-risk items, items needing a decision, and AI in use by day across Microsoft 365, browser and device.
The Centeye Lumen Discovery screen for a test client. Your check report is built from the same data.

What the check finds

Six things every business should know about its AI.

01

AI tools in use, and by whom

Which AI tools people sign in to and which AI apps they have connected, with a register entry and an owner for each.

02

AI app access left open

Every AI app connected to mail, files or calendars, what each grant allows, and the ones nobody has used in months.

03

AI agents and their identities

Agents built in Copilot Studio, and Copilot and Gemini agents. It flags any agent running on a shared admin account or a person’s own login.

04

What Copilot could surface

Folders shared with everyone, such as payroll or HR, that an AI assistant can read out to anyone who asks.

05

Data going into AI tools

Add the Lumen browser extension on a few devices to see what gets pasted or uploaded into AI tools, and how much, including on personal accounts. It never reads the content.

06

MCP server connections

Which MCP servers people’s AI tools connect to, what each one can reach, and which nobody approved, so you can hold or block them.

What it looks like in Lumen

AI use from three sides: Microsoft 365, the browser and the device.

Real Centeye Lumen charts from a test client. Your check is built from the same views.

Lumen chart: distinct AI services in use each day for 14 days, with separate lines for Microsoft 365, browser and device.
AI in use, by day. Distinct AI services seen each day, by source. Captured early on 4 Nov, so that day's lines start at zero.
Lumen risk panel: 3 high, 3 medium, 2 low and 50 not read, out of 58 AI items.
Risk, from reach. What each AI item can reach. Items not read yet show as Not read, never as low risk.
Lumen chart: items per AI service such as ChatGPT, Copilot, Claude and Otter, split into Microsoft 365 app grants, browser sites and device programs.
Footprint by service. Each AI service, split by where Lumen saw it and how many people use it. Browser and device data need the optional Lumen browser extension and device sensor.

The report

One list. Riskiest first.

A plain-language report for you and whoever runs your IT.

AI exposure · Harbor Lane Dental Sample data
WhatTypeOwnerCan reachRiskFix
Meeting notes appConnected AI appLeft the companyEvery mailbox and calendar HighRevoke, or name a new owner and narrow it
Invoice helperCopilot Studio agentNone (shared admin account)All finance files HighGive it its own identity, limit it to the invoices folder
ChatGPTAI tool · 14 peopleNot approvedClient files pasted in, 5 on personal accounts* MediumApprove a work account and coach people to use it
HR questions agentCopilot Studio agentHR managerPayroll folder (shared with everyone) MediumFix the folder’s sharing
Proposal drafterGeminiOffice managerShared drives it needs LowSet a review date
Made-up business, to show the layout of the report. *Paste and personal-account data come from the optional Lumen browser extension.

How it works

Three steps. Nothing changes until you say so.

  1. 1RequestAnswer one quick question about your company.
  2. 2ConnectYour Microsoft 365 or Google Workspace admin approves Centeye. Centeye starts listing the AI apps and agents already there within minutes.
  3. 3ReportYou get the ranked list, with the fix for each risk. Walk through it with us on a short call if you like.

Honest scope. The check covers Microsoft 365, Microsoft Entra ID, Google Workspace and the AI apps connected to them. Seeing what is pasted into AI tools needs the optional Lumen browser extension. It is a one-time check, not ongoing protection: Centeye Synthwatch and Centeye Lumen do the watching afterwards, if you want them to.

Questions

The AI Exposure Check, answered.

Is the AI Exposure Check really free?
Yes. The check is free, with no credit card. If you want Centeye to keep watching after the report, we talk about that once you’ve seen it.
Does the check change anything in the tenant?
No. Everything in Centeye starts on “tell me”: it shows you what it finds and changes nothing until you say so.
What do you need from us?
Your Microsoft 365 or Google Workspace admin approves Centeye with one sign-in. Nothing needs installing for that. To also see what gets pasted into AI tools, add the Lumen browser extension on a few devices.
What does the AI Exposure Check look at?
The AI tools people use and the AI apps connected to Microsoft 365 or Google Workspace, the agents built in Copilot Studio, Copilot and Gemini agents, and files shared so widely an AI assistant could surface them. Centeye finds agents through the identities and access they hold, so it works whichever AI vendor built them.
Is this the same as a shadow AI audit?
It covers the same ground in plain terms: which AI is in use, who uses it, what it can reach and what to fix first. It is a check, not a certification, and it doesn’t replace your own policies.
What about the people’s sign-ins?
That’s the Identity Exposure Check, our free check of sign-ins and accounts in Microsoft 365 or Google Workspace. Many teams run both.

Read more: Shadow AI detection · AI data leaks · AI agent identity · Copilot agent security · How to secure AI agents · Copilot security · Identity Exposure Check · An MSP or IT provider? Check a client instead

Microsoft, Microsoft 365, Copilot, Copilot Studio and Entra are trademarks of the Microsoft group of companies. Google Workspace and Gemini are trademarks of Google LLC. Use here describes compatibility only.

Find out where your company data goes in AI.

Free, read-only to start, and no credit card.

Request your free check