ChatGPT: Keep Client Data Safe, Set Up Your Team’s Data Controls in 10 Minutes

Why this matters

A client emails before renewing: "Can you confirm whether our files are ever put into ChatGPT, and what settings you use?" Their insurer now asks every vendor. You need a short written answer, and right now you do not have one.

Worse, you know that last week a new hire pasted a client payroll export into ChatGPT on a personal login, and nobody knows which settings that account has. No bad intent, no policy, default settings. That is how most small firms use AI today.

In this module you fix it in about ten minutes: set the controls that matter, learn what export and delete are for, and leave with a one-page team AI policy you can send to the client the same day.

What you’ll build

You are Priya, office manager at Hartwell Bookkeeping, a made-up six-person firm. Starting from the client’s email, you switch off model training, review shared links, find Export data and Delete all chats, decide on memory, and try Temporary Chat.

You finish with Hartwell’s one-page team AI policy: one approved account, training off, memory off on shared logins, Temporary Chat for sensitive work, a never-paste list, a monthly shared-links check, export before delete, and one named person to ask.

What you need

Nothing to start the practiceThe inbox, ChatGPT and the policy are practice copies. No setting really changes and the firm and client are made up.
The ChatGPT account your team uses for workchatgpt.com. Do this on every account anyone uses for client work. On Business or Enterprise, the workspace admin controls these settings.
Where each setting livesClick your name at the bottom left, then Settings. Training, shared links, export and delete are under Data controls. Memory is under Personalization. Temporary Chat is the icon at the top right of a chat.
Your client contractsCheck what you have promised clients about their data. Your contracts and local rules come before this policy. This is not legal advice.

Concepts

4 ideas you’ll use in the practice.

Training and shared links: where chats can travel

Improve the model for everyone decides whether your chats can be used to train OpenAI’s models. Switch it off on personal plans. OpenAI says Business and Enterprise plans do not train on business data by default.

Shared links are the other leak. Anyone with a shared link can read that chat, and a link to a client conversation can sit there for months. Settings, Data controls, Shared links, Manage lists them. Check it monthly.

Memory: useful alone, risky when shared

Memory lets ChatGPT carry details from one chat into the next. On your own login that saves typing. On a login the whole team shares, it can bring one client’s details into another client’s chat.

Most firms switch memory off on shared accounts and, on any account that keeps it, look at the Memory summary now and then to see what it has kept.

Temporary Chat: safer, not secret

A Temporary Chat ignores memory and does not appear in your history. Use it for the one sensitive job you cannot avoid.

ChatGPT’s own notice says a copy may still be kept for up to 30 days for safety. So names and account numbers still come out first.

A policy is what makes settings stick

Settings only protect you if everyone uses the same account with the same settings. One page of written rules, with one person who owns them, is what turns ten minutes of setup into a team habit, and into an answer you can send a client.

Practice

Do every step yourself in a practice copy of the real tool. It checks your work, and “Do it for me” finishes any step you get stuck on.

ChatGPTChatGPT: Keep Client Data Safe, Set Up Your Team’s Data Controls in 10 Minutes16 steps · ~10 min · nothing touches your real accounts

Start the practice →Prefer to read? Step-by-step guide

Do it for real

Now repeat it in your own account. Tick each step as you go.

List every ChatGPT account anyone uses for client work. Pick one approved account (a business plan if you can).
On each account: your name, Settings, Data controls. Switch off Improve the model for everyone.
On the same page, open Shared links, Manage, and delete any link nobody needs.
Under Settings, Personalization, switch off memory on any login more than one person uses.
Try Temporary Chat from the icon at the top right, so you know where it is for sensitive work.
Copy the policy below, change the firm name, the rules and the owner to fit you, and send it to the team.
Answer the client in writing: which account you use, which settings are on or off, and what your team never pastes.

Downloads

One-page team AI policy (.txt)Hartwell’s policy from the end of the lesson. Replace the firm name, the rules and the owner with yours.

Quick check

3 questions to check you’ve got it.

1. A client asks if their files train ChatGPT. Which setting answers that?

Memory
Improve the model for everyone, under Data controls
Temporary Chat

Show the answer

Improve the model for everyone, under Data controls. That switch decides whether chats may be used for training. Memory and Temporary Chat are about what ChatGPT carries between chats.

2. Why do most firms switch memory off on a shared login?

It makes ChatGPT slower
One client’s details could turn up in another client’s chat
Memory costs extra

Show the answer

One client’s details could turn up in another client’s chat. Memory carries details from chat to chat. On a shared account, those can be another client’s.

3. Is a Temporary Chat completely private?

Yes, nothing is ever kept
No, a copy may be kept for up to 30 days, so take out private details first
Only on weekends

Show the answer

No, a copy may be kept for up to 30 days, so take out private details first. It stays out of your history and ignores memory, but it is safer, not secret.

Up next · Module 3

Roll Out a Team AI Policy in One Afternoon using Claude

Your account is set up. Next, write the rules for the whole team: Claude turns your decisions into a one-page AI policy, an announcement email and a five-question staff quiz.

Start module 3 →