You have a client’s contract to summarise, a messy customer list to clean up, or an angry email to answer. Pasting it into ChatGPT would take ten seconds. But is it safe?
The short answer: it depends on what you paste, which plan you use and how your settings are set. Pasting a client’s private data into a free ChatGPT account with default settings is risky: your chats may be used to train future models, and they’re stored on someone else’s servers. Pasting the same task with names and numbers removed, or using a business plan that doesn’t train on your data, is usually fine.
This guide explains what really happens to what you paste into ChatGPT, Gemini and Claude, the risks that matter for a small business, and simple rules to follow, so you can use AI without breaking your clients’ trust.
What actually happens to what you paste
When you paste something into an AI chat, three things matter: whether it’s used to train future models, how long it’s kept, and whether a person might read it. The answers depend on the tool and your plan.
| Tool and plan | Used for training by default? | How long chats are kept |
| ChatGPT Free, Plus and Pro | Yes, unless you turn it off | Until you delete them; deleted chats are removed within 30 days |
| ChatGPT Business, Enterprise and API | No | Set by the workspace and contract |
| Gemini (personal Google account) | Yes, while “Keep Activity” is on | 18 months by default; 72 hours with Keep Activity off; chats seen by reviewers up to 3 years |
| Gemini in Google Workspace (business) | Covered by Workspace’s separate data terms | Set by your Workspace admin |
| Claude Free, Pro and Max | Only if you allow it in Privacy Settings | 30 days, or up to 5 years if you allow training |
| Claude Team, Enterprise and API | No | Set by the contract |
Checked against OpenAI, Google and Anthropic’s own policy pages, October 2026. Policies change, so check again before you rely on them.
A few details worth knowing:
- Training isn’t the same as storing. Even with training turned off, your chats are kept on the provider’s servers for a while for safety and abuse checks. Turning training off doesn’t make a chat disappear.
- People can read some chats. Google says a sample of Gemini chats is reviewed by humans, even with Keep Activity off for safety reasons, and its help page asks users not to enter confidential information they wouldn’t want a reviewer to see.
- Courts can order data to be kept. In 2025, a US court order in the New York Times lawsuit required OpenAI to keep even deleted ChatGPT chats for several months. That order ended in September 2025, but it showed that “deleted” doesn’t always mean gone right away.
- Temporary chats help. ChatGPT’s Temporary Chat and Gemini’s temporary chats aren’t used for training and don’t appear in your history, which makes them a better choice for one-off sensitive tasks.
The real risks for a small business
The chance that ChatGPT will repeat your client’s exact details to a stranger is low. The bigger risks are more ordinary, and they’re the ones that cost clients.
1. Breaking your contract or NDA
Many client contracts and NDAs say you can’t share their information with third parties without permission. An AI provider is a third party. If a client finds out you pasted their confidential plans into a free AI tool, you may have broken the agreement, even if nothing leaked. Read your contracts, or ask clients directly whether they’re comfortable with AI tools.
2. Privacy laws
Personal data, meaning anything that identifies a person like names, emails, phone numbers and addresses, is protected by law in many places: GDPR in the UK and EU, India’s Digital Personal Data Protection Act, and various US state laws, with stricter rules for health and financial data. Pasting customers’ personal details into a tool without a proper agreement can put you on the wrong side of these laws. Business plans usually come with a data processing agreement. Free accounts don’t.
3. Staff pasting things you didn’t know about
In 2023, Samsung restricted staff use of AI chatbots after employees pasted confidential source code into ChatGPT. Most leaks like this aren’t hacks. They’re well-meaning people trying to work faster. Without a simple rule, your team will make its own decisions.
4. Your AI account getting hacked
Your chat history is a record of everything you’ve pasted. If someone gets into your ChatGPT account, they can read it all. Protect AI accounts like email: a strong unique password and two-factor authentication.
5. Shared links
ChatGPT and other tools let you share a chat as a link. Anyone with the link can read it. In 2025, some shared ChatGPT chats even appeared in Google search results through an opt-in feature that OpenAI then removed. Don’t share chats that contain client details, and delete old shared links you no longer need.
What’s usually fine, and what to never paste
A simple traffic-light rule works for most small businesses:
| Usually fine | Only after anonymising | Never paste |
| Your own drafts, ideas and marketing copy | Client emails and complaints | Passwords, API keys and login links |
| Public information, like a client’s published website | Contracts and proposals | Card numbers and bank details |
| General questions about your industry | Meeting notes and project details | ID documents, tax IDs and passport scans |
| Templates with placeholders instead of real names | Spreadsheets of customer or sales data | Health, legal or HR records about real people |
How to anonymise in two minutes
The AI almost never needs to know who someone is. It only needs the situation. Before you paste, swap the details that identify people or companies:
- Replace names with roles or placeholders: “Priya at Acme Ltd” becomes “[Client], a retail company”.
- Remove contact details: emails, phone numbers and addresses.
- Round or replace numbers that could identify a deal: “₹4,37,500” becomes “about ₹4.4 lakh”, or “[amount]”.
- Strip out IDs: invoice numbers, account numbers and order numbers, unless the task truly needs them.
- Paste only the part you need. Summarising one clause? Paste that clause, not the whole contract.
Then put the real details back into the AI’s draft yourself. For example, instead of pasting a full client email, write: “A client is upset that their website launch was delayed by a week because they sent content late. Write a calm reply that explains the delay and offers a new launch date.” You get the same quality of draft with nothing private shared. Our ChatGPT prompts for business emails are written this way, with placeholders in square brackets.
Settings to change today
These take about five minutes and work on free and personal paid plans.
ChatGPT
- Go to Settings → Data controls and turn off “Improve the model for everyone”.
- In the same menu, open Shared links and delete any you no longer need.
- For one-off sensitive tasks, start a Temporary Chat from the top of the chat screen.
Gemini
- Open Gemini Apps Activity in your Google account settings and turn Keep Activity off, or set auto-delete to the shortest period.
- Use temporary chats for anything sensitive.
Claude
- Go to Settings → Privacy and check the model training setting. Leave it off if you don’t want your chats used for training.
All of them
- Use a strong, unique password and turn on two-factor authentication. A password manager makes this easy.
When to pay for a business plan
Settings reduce the risk, but a business plan changes the terms. ChatGPT Business, Claude Team and Gemini in Google Workspace don’t use your data for training by default, give you admin controls over who uses what, and come with data processing terms you can show a client. Consider upgrading when:
- you regularly work with client documents or customer data;
- two or more people in your business use AI tools;
- a client asks how you handle their data, or your contracts mention data protection.
Not sure which tool suits you? Our comparison of ChatGPT vs Claude vs Gemini covers how they differ for everyday business work.
A five-rule AI policy for your team
Even a two-person business benefits from writing this down. Copy it, adjust it, and share it:
- Use only these AI tools for work: [list the tools and plans you’ve approved].
- Never paste passwords, card or bank details, ID documents, or health, legal or HR records about real people.
- Remove names, contact details and account numbers from client material before pasting it. Use placeholders.
- Check every AI output for mistakes before it goes to a client. You are responsible for what you send.
- If you pasted something you shouldn’t have, delete the chat and tell [name] the same day. No blame, just fix it.
Related guides: AI for Small Business: A Practical Guide to Tools, Automation and Privacy · Small Business Cybersecurity: The Complete Guide
FAQ
Does ChatGPT keep what I type?
Yes, for a while. On personal plans, chats stay in your history until you delete them, and deleted chats are removed from OpenAI’s systems within 30 days, except where the law or security requires otherwise. Temporary Chats don’t appear in your history and aren’t used for training.
If I turn off training, is my data private?
It’s more private, not fully private. Your chats are still stored for a period and can be checked for abuse, and the provider still processes them to answer you. For client data, combine the setting with anonymising, or use a business plan.
Is ChatGPT Business or Claude Team safe for client data?
They’re much safer than free accounts: they don’t train on your data by default, and they offer admin controls and data processing terms. They’re still a third party, so check your client contracts and avoid pasting the most sensitive records unless you really need to.
Can I use AI to summarise a client contract?
Yes, with care. Remove the client’s name, addresses and anything identifying, or paste only the clauses you need help with. And treat the summary as a starting point, not legal advice.
What should I do if I already pasted something sensitive?
Delete the chat, and delete any shared link to it. If it was a password or a key, change it immediately. If it was a client’s personal or confidential data, check your contract and local privacy law, which may require you to tell the client.
Final word
AI tools are safe enough for most small business work, as long as you treat them like any outside service. Turn off training on personal accounts, anonymise client material before pasting, and never paste passwords, financial details or sensitive personal records. When AI becomes part of your daily client work, move to a business plan and write the five-rule policy down for your team.
The simplest test before you paste: would your client be comfortable seeing this in someone else’s system? If not, change the details first.


