Plain-English guide
Can you put customer information
into ChatGPT?
This is the question behind the most common real-world data leak in small business. Not hackers, not ransomware — someone in the office pasting a client's details into a free AI tool to save twenty minutes. Here is when that is fine, when it is a problem, and how to sort it without banning anything.
Putting customer information into ChatGPT is not automatically illegal, but it breaches UK GDPR if you have no lawful basis, your privacy notice does not mention it, or the data is special-category such as health details. The critical distinction is account tier: consumer plans may use your conversations for training unless you opt out, while business and enterprise plans exclude them by default.
Published
The short answer
No, it is not automatically illegal. Anyone who tells you it is has not read the law. UK GDPR does not ban AI tools; it governs how you handle personal data, whatever you handle it with.
It becomes a problem in four situations: you have no lawful basis for the processing, your privacy notice does not mention AI at all, the tool may use that data to train a model, or the information is special-category data that should not be in a general AI tool under any circumstances.
Most small firms fall foul of the middle two without ever making a decision about it. Nobody sat down and chose to leak client data. Someone was busy, the tool was open, and the paperwork nobody had written did not stop them.
The mechanics
What actually happens to what you type
People imagine one of two extremes: either everything typed into ChatGPT is instantly public, or none of it goes anywhere. Neither is right, and the truth depends on something most users have never checked — which kind of account they are logged into.
On consumer plans — the free tier and the individual paid tiers — your conversations may be used to help improve the models unless you go into the data controls and switch that off. It is a setting, and it is on by default.
On business plans and the API, your content is excluded from training by default. OpenAI states this plainly for ChatGPT Business, Enterprise and the API platform: there is nothing to opt out of, because it was never opted in. Business tiers also come with a data processing agreement, which matters because it is the document that makes your provider a processor acting on your instructions rather than someone doing whatever they like with your customers' details.
Worth knowing: ChatGPT Team was renamed ChatGPT Business in August 2025. It is the same plan. If you are reading advice that still says "Team", check the rest of it for age too.
The one that catches people
Paying is not the same as protecting
This is the misunderstanding I run into most often, and it catches sensible people. Someone upgrades to a paid individual plan, reasonably assumes that paying customers get better treatment of their data, and carries on pasting client details in.
The dividing line is not free versus paid. It is consumer tier versus business tier. An individual paid plan buys you better models and higher usage limits. It does not buy you different data handling, an admin console, or a data processing agreement.
The same pattern holds across the main providers, not just OpenAI. Individual paid plans on other assistants are generally treated as consumer accounts too, with training controls you have to find and change yourself. If you are using AI for client work on a personal account, go into the settings today and check what the training toggle says. It takes two minutes and most people are surprised.
When it is fine, and when it is not
Generally fine
- Drafting a reply on a business-tier account, where the customer's name is the only personal detail and your privacy notice covers AI use.
- Working on anything with the names and identifying details stripped out first. Redaction is underrated and takes seconds.
- Anything with no personal data in it at all: pricing structures, method statements, marketing copy, your own processes.
Not fine
- A consumer account with training left on, handling identifiable customer information.
- Anything special-category — health, ethnicity, religion, sexual orientation, biometrics — regardless of tier.
- A client's data that you hold under contract, where your agreement with them says nothing about AI processing.
- Any of the above where your privacy notice makes no mention of AI. The processing might be defensible; the silence is not.
The ICO's guidance on AI and data protection, on ico.org.uk, is the fuller reference if you want it. The headline is that AI does not get its own special rules — the ordinary principles apply, and you have to be able to show which lawful basis covers what you are doing.
What should never go in
There is a short list where the answer is no regardless of tier, settings, or how careful you are being. Special-category data carries a much higher bar under UK GDPR and a general-purpose AI tool is not the place for it:
Health and medical information. Ethnic or racial origin. Religious or philosophical beliefs. Sexual orientation. Biometric or genetic data. Trade union membership. Political opinions. Also, though not special-category: anything covered by a confidentiality agreement, and anything belonging to a client rather than to you.
This is the same reasoning behind my own scoping: I do not take on work involving medical records, and I would tell you the same about your own tooling. Some data needs a specialist environment, not a clever general assistant.
How to make it safe in an afternoon
Four things, and none of them require banning AI, which never works anyway. Staff who find a tool useful will use it on their phones instead, and then you have the same exposure with none of the visibility.
- 01
Move daily users onto a business tier. This is the single highest-value change. Training is off by default, you get an admin console, and you get a data processing agreement. For a handful of seats it is a modest monthly cost against the alternative.
- 02
Check the settings on anything that stays consumer. Anyone using a personal account for work needs the training toggle turned off today.
- 03
Add a paragraph to your privacy notice. That you use AI tools to help run the business, roughly what for, and that customer data is not used to train public models. Short and honest.
- 04
Write the one-page staff policy. Which accounts to use, what may never go in, and who to ask when it is not obvious. This is what actually changes behaviour.
If it has already been happening
It probably has. In most firms I look at, someone has been using a free account for real work for months and nobody knew, because nobody asked and there was no rule to break.
Do not run a witch hunt. You will find out far more by asking openly what people have found useful than by making it a disciplinary matter, and the person who tells you is doing you a favour. Then: establish roughly what has gone in and how sensitive it was, move everyone onto proper accounts, turn training off where it was on, and get the privacy notice and policy written.
Whether anything is reportable to the ICO depends on what was exposed and the risk to the people involved. Routine business correspondence going into a tool with training enabled is untidy rather than catastrophic. Special-category data is a different conversation, and that is the point at which you want a data-protection specialist rather than a blog post. I will tell you straight if you have crossed that line.
Where this comes from
This is plain-English guidance for small businesses, not formal legal advice. Where something genuinely needs a solicitor or a data-protection specialist, I'll tell you straight.
Straight answers
- Is it illegal to put customer information into ChatGPT?
- Not automatically. It becomes a UK GDPR problem when you have no lawful basis for it, when your privacy notice does not mention it, when the tool may use that data to train a model, or when the information is special-category data such as health details. Everyday drafting with a business-tier account and no sensitive data is usually fine.
- Does ChatGPT train on what I type?
- It depends entirely on which account you are using. Consumer plans — Free, Go, Plus and Pro — may use your conversations to improve the models unless you turn that off in the data controls. Business and Enterprise plans and the API are excluded from training by default, with nothing to opt out of.
- I pay for ChatGPT Plus, so my data is protected?
- No, and this is the most common misunderstanding I come across. Plus and Pro are consumer plans. Paying more gets you better models and higher limits, not different data handling. The dividing line is consumer versus business tier, not free versus paid.
- What counts as customer information?
- Anything that identifies a living person. A name, an email address, a phone number, an address, a job reference tied to a customer, a complaint, a photo of someone. It is a much wider net than most people assume — a quote with a name and address on it is personal data.
- What must I never put in?
- Special-category data: health information, ethnicity, religion, sexual orientation, biometrics, trade union membership. Also anything under a confidentiality agreement, and anything belonging to a client rather than to you. These do not belong in a general AI tool at all, whatever tier you are on.
- Do I have to tell customers I use AI on their data?
- Yes. If personal data goes into an AI tool, your privacy notice has to say so and roughly what for. Silent processing is the breach people actually get caught by, and it is one paragraph to fix.
- What if my staff have already been doing this?
- Do not panic and do not start a witch hunt. Find out what has gone in, switch the accounts to a business tier, turn training off where it applies, update your privacy notice, and write a short policy so it does not continue. Whether it is reportable depends on what was exposed and how sensitive it was.
- Is a UK-hosted AI tool safer?
- It removes one question, not all of them. International transfers have their own UK GDPR rules and most mainstream providers now offer data residency or appropriate safeguards. Where the data sits matters less than whether it is used for training and whether you told anyone you were doing it.
Read next
Does the EU AI Act apply to my UK business?
It came into force on 2 August 2026 and being in Cornwall does not put you outside it. Who it catches, what it asks for, and the December deadline.
Do I have to tell customers they are talking to a bot?
No single UK law says so. EU law now does. Where the line sits, and why one sentence solves it either way.
Want this sorted properly?
A fixed-price AI-readiness check covers the lot: which accounts your staff are actually using, the privacy notice wording, a one-page policy, and an honest view of what is already exposed. Plain English, no open-ended bills.
Ask about a readiness check