Microsoft Copilot
Answers from SharePoint, OneDrive, Teams and Outlook, using whatever the signed-in user can already open.
AI data exposure · New York · New Jersey · Worldwide
Your AI tools can see more than you think.
Copilot, Gemini and connected ChatGPT or Claude workspaces answer from every file a user can open. Staff also paste company data into personal AI accounts. Pick a question and look through the lens.
Market avg. $4,950. Your fixed quote: under it.
Simulated example · file names are illustrative
Where AI reads from
What can AI assistants see in your company?
An AI assistant connected to company files can see everything the person using it is allowed to open. It does not get extra access, and it does not check whether that access still makes sense. Here is how that plays out on each platform.
Google Gemini
Reads Drive, Docs and Gmail inside Workspace, with the same reach as the person asking.
ChatGPT and Claude
Business plans can connect to SharePoint, OneDrive and Google Drive. Once connected, they search those files on the user’s behalf.
Personal AI accounts
A free account on any of these tools sits outside your control. Whatever staff paste in stays with that account.
The pattern we find
How do private files end up in AI answers?
They were shared too widely years ago and nobody noticed, because finding them used to take effort. AI assistants remove that effort. Salary files, board material, HR cases and client contracts turn up most often.
- 1
A project starts
Someone creates a SharePoint site or a shared Drive folder for a new client.
- 2
Sharing goes wide
Picking names takes time, so it gets shared with the whole company.
- 3
The project ends
The files stay where they were, open to everyone, and nobody looks again.
- 4
Someone asks AI
Years later, one plain-English question pulls up the salary spreadsheet.
The audit
What does the AI data exposure audit cover?
It covers five layers, in the order that matters most: what is over-shared, what is sensitive, what should be blocked from leaving, which AI tools are already in use, and what your staff are allowed to do with them.
- 01
Permission exposure
SharePoint, OneDrive, Teams and Google Drive reviewed for company-wide links, anonymous links and guests who never left.
- 02
Sensitive content
Payroll, HR, board and client files found and labeled, with Microsoft Purview or Google Drive labels applied where the pattern is reliable.
- 03
Data loss prevention
Rules that stop labeled files from leaving, tested against real documents from your environment.
- 04
Shadow AI discovery
Which AI tools and personal accounts staff already use for work, and what kind of data goes into them.
- 05
AI use policy
A plain-language policy your staff will actually read, covering which tools are approved and what never goes into them.
Two-minute self-check
How exposed is your company right now?
Five yes-or-no questions give you a rough exposure score. Your answers stay in your browser unless you choose to send them with a review request.
Beyond the audit
What else can Kildaris help with?
Once your data is locked down, we help you get real work out of AI on ChatGPT, Claude, Gemini, Copilot or any other AI platform, and we run the IT underneath it. Every quote comes in under the typical market cost. See how we compare.
market avg. $3,500Your fixed quote comes in under it
AI assessment
Find the work AI can take off your team, on any platform.
Details market avg. $5,000Your fixed quote comes in under itWorkflow automation
One busywork process automated, with a person approving what matters.
Details market $8,950Your fixed quote comes in under itCopilot adoption
Get the seats you already pay for into daily use.
Details from $75/user/moManaged IT
Help desk and monitoring, 24x7x365, with patching, backup and security as quoted add-ons.
DetailsQuestions
Common questions about AI data exposure
Can Microsoft Copilot show people files they should not see?
Copilot only returns content the signed-in user already has permission to open. The risk is that permissions in most Microsoft 365 tenants are wider than anyone intended: links shared with the whole organization, "anyone with the link" links, and guests from old projects. Copilot makes those files easy to find with a plain question.
Does the same risk apply to Gemini, ChatGPT and Claude?
Yes, whenever they are connected to company files. Gemini in Google Workspace answers from Drive content the user can open, and ChatGPT or Claude connected to SharePoint, OneDrive or Google Drive work the same way. Personal AI accounts are a different problem: data pasted into them leaves your control entirely.
What does the AI data exposure audit cost?
A typical published fixed fee for an AI governance audit is about $4,950. Your Kildaris quote is a fixed price after a free 30-minute call, set to come in under the typical market cost for the same scope. We agree the scope in writing before any work starts, so there are no hourly surprises.
How do you find staff using personal AI accounts?
We combine sign-in logs, DNS and network records, browser extension inventory and an expense review. The expense review usually finds the most, because unapproved AI tools tend to be paid for on a company card.
Do we have to switch off our AI tools during the audit?
No. Your team keeps working while we review. If we find something urgent, such as a payroll file open to anyone with the link, we tell you the same day.
Which areas do you cover?
We are based in the New York and New Jersey metro and work on site across the five boroughs, Long Island, Westchester and New Jersey, so Long Island is covered. Most of the audit is done remotely, so we also work with teams worldwide.
Next step
Audit it before AI finds it.
Thirty minutes to scope the review, no obligation. We look at which AI tools you use and how your files are shared, then quote a fixed fee.
Or email info@kildaris.com