A plan for a small firm

How Do I Stop Employees Pasting Client Data Into ChatGPT?

Updated November 16, 2026

Short answer: You won't stop it with a rule alone, and you won't stop it with a block alone. Staff paste client data into ChatGPT because it saves them time. The plan that works in a small firm has five parts: find out what people already use, give them an approved AI tool whose terms you have read, write down what client data may and may not go in, switch off training and set retention where the tool allows it, and check in every few months. Blocking personal AI sites on work devices can back this up, but it only works once staff have something better to use.

Why do staff do it?

Because it works. IBM's explanation of shadow AI is that "Employees often turn to these tools to enhance productivity and expedite processes" (IBM). A staff member summarizing a 40-page policy document or drafting a client letter is trying to do the job faster.

They also often don't know the rules. In a CybSafe and National Cybersecurity Alliance survey, 52% of employed participants said they had not received any training on safe AI use (Infosecurity Magazine). If nobody has told them what is allowed, they will guess.

Step 1: Find out what is already happening

Ask your team, and promise that nobody is in trouble for past use. You want to know:

  • Which AI tools they use, and whether on a personal or firm account.
  • What they use them for. Drafting? Summarizing client files? Checking numbers?
  • What kinds of client information have gone in.

This tells you what the approved tool must be able to do. If half the team uses AI to summarize long documents, an approved tool that can't read files won't replace anything.

Step 2: Give them an approved tool

This is the step that matters most. A rule that says "no AI" competes with a tool that saves an hour a day, and the tool often wins.

An approved tool should be a business plan your firm owns, not a collection of personal accounts. Before you approve one, read its own pages for four things: whether it trains on your data, how long it keeps chats and files, who at the vendor can see them, and whether you can add and remove users. What makes an AI tool safe to approve sets out how ChatGPT Business, Claude Team, Microsoft 365 Copilot and Faradex answer those questions.

Step 3: Write down what may go in

A one-page policy is enough for most small firms. The part staff will actually use is a short list of what may and may not go into AI. For example:

Kind of information Personal AI account Approved firm tool
Public information, your own marketing copy Allowed Allowed
Internal notes with no client details Not allowed Allowed
Client names, account numbers, policy or tax details Not allowed Allowed, if your policy and the client's contract permit it
Anything a client contract or NDA says must not be shared Not allowed Only if the contract permits it

Use the nouns your staff actually work with: census files and claims histories in an agency, client statements and financial plans for an adviser, returns and payroll for an accounting firm, contracts and candidate files in professional services. Our AI acceptable use policy template has a fuller version you can copy.

Step 4: Set the tool up on purpose

Most business plans let an admin change settings that matter:

  • Training. Check the default. ChatGPT Business says "By default, we do not use your business data for training our models" (OpenAI). Personal ChatGPT plans have data sharing "enabled by default" (OpenAI).
  • Retention. Where the plan lets your admin choose how long chats are kept, choose a period that fits your own record-keeping instead of leaving the default.
  • Who has access. Add only current staff, and remove people the day they leave.
  • Sign-in. Where the plan offers single sign-on, use it so that leaving the firm also ends AI access.

Step 5: Back it up, and check in

Once staff have an approved tool, you can reasonably ask them to stop using personal accounts for client work, and you can back that up:

  • Ask staff to confirm in writing that they have read the policy.
  • Consider blocking personal AI sites on firm devices. The blocking page explains what that does and does not cover.
  • Every three to six months, ask again what people use. New tools appear fast.

What not to do

  • Don't rely on a ban with no alternative. Samsung restricted generative AI on company devices after internal data was leaked to ChatGPT, and at the same time said it was "reviewing measures to create a secure environment for safely using generative AI" (TechCrunch). Even a company that size treated the ban as temporary.
  • Don't buy monitoring software before you have a policy. You will be measuring people against rules they have never seen.
  • Don't assume "no training" means "not kept". A business plan that doesn't train on your data can still keep every chat for as long as your settings allow.

About Faradex. Faradex is a private AI workspace for small and mid-sized firms that handle confidential client work. Each firm gets its own Workspace on dedicated infrastructure in the United States. Files and conversations are held in memory and are gone when the Workspace is reset. AI model: Currently Opus via AWS Bedrock, more information is available on our website. AWS does not store prompts or responses, and model providers have no access to them. Pricing: from $120 a month, no seat minimum; see faradex.ai/pricing.