Is ChatGPT safe for business? A UK guide to customer data

I'm Managing Partner of a two-office law firm in East Sussex with around 30 people, so client confidentiality is part of my working day. I built the firm's AI-assisted marketing workflow, which runs automatic checks before a person approves anything.

Is ChatGPT safe to use for business?

It can be, once it's set up properly. The everyday risks are less about hackers and more about what staff paste in, which account they use, and whether anyone checks the answer before it reaches a customer. Get those three right and ChatGPT becomes a sensible business tool, not a leak waiting to happen.

The National Cyber Security Centre (NCSC) made the core point back in 2023. Whatever you type is visible to the company running the tool, and it's stored. Its advice was to keep sensitive information out of public AI tools, along with anything that would cause trouble if it became public.

For a small business, the risks fall into three areas:

• Data: where your customers' information goes, who can see it, and whether it trains the model.
• Accuracy: ChatGPT can be wrong and sound certain. OpenAI says so itself.
• People: staff on personal accounts, with no rules and no record of what went in.

Is ChatGPT GDPR compliant?

No tool is GDPR compliant on its own. Compliance depends on how your business uses it. OpenAI offers business plans with a data processing agreement and no training on your data by default, which helps. But you remain responsible for what goes in, why it goes in, and how long it's kept.

The ICO has been clear that there's no “AI exemption” from data protection law. If staff put personal data into ChatGPT, UK GDPR and the Data Protection Act 2018 apply, as they would to any other system. Personal data means anything that identifies a person directly or indirectly, so it's wider than names and email addresses.

On a business plan, your business is normally the controller and OpenAI acts as your processor. The ICO says a controller using a processor must have a written contract with it. For ChatGPT Business, Enterprise and OpenAI's API, that contract is OpenAI's data processing addendum.

Under that addendum, UK data is handled by OpenAI's US company, using the Information Commissioner's UK addendum to the EU standard contractual clauses. A personal account is different. It's an agreement between your employee and OpenAI, and your business isn't a party to it. So there's no processor contract covering your customers' data.

How does ChatGPT data privacy differ between personal and business accounts?

Personal plans (Free, Go, Plus and Pro) belong to the individual, and chats can train OpenAI's models unless the user opts out. ChatGPT Business and Enterprise belong to your company. Training is off by default, an admin controls members and retention, and OpenAI will sign a data processing agreement with you.

Does ChatGPT train on what you type?

On personal plans, yes, unless the user switches off “Improve the model for everyone” in Data controls. On ChatGPT Business and Enterprise, OpenAI says it doesn't train on your data by default. Either way, your chats are still stored by OpenAI, so a training setting alone doesn't make customer data safe to paste.

OpenAI explains the detail in its help article on how your data is used to improve model performance. A few points matter for ChatGPT data protection in a business:

• Feedback still counts. Even with training switched off, a thumbs up or down on an answer means that whole conversation may be used for training.
• Temporary Chat helps a little. Temporary chats stay out of your history, aren't used for training and are deleted after 30 days. They still go to OpenAI, so they're no workaround for customer data.
• Deleting takes time. Deleted chats are scheduled for permanent deletion within 30 days, unless OpenAI must keep them longer for security or legal reasons.
• Connected apps widen the reach. ChatGPT only sees your files or inbox if someone uploads them or connects an app such as SharePoint, Outlook or Google Drive. On a Business workspace, an admin can switch those apps off.

What should staff never paste into ChatGPT?

Never paste special category data, such as someone's health or religion. The same goes for criminal records, confidential client files, passwords and bank details. Customer names and contact details belong in a business account only, and only when the job needs them. If you'd hate to see it in a newspaper, keep it out.

Taking a name out isn't always enough. If someone could still work out who the person is, it's still personal data. The principle of data minimisation applies too: use only the personal data the job needs.

Even OpenAI's own business contract says no sensitive data is intended to be sent, unless a user includes it unexpectedly. That tells you how the tool is meant to be used.

Do you need a DPIA to use ChatGPT?

Possibly. The ICO says you must do a data protection impact assessment when processing is likely to be high risk. It treats AI as innovative technology, and requires a DPIA when that's combined with another factor, such as sensitive data or processing on a large scale. If you're unsure, the ICO recommends doing one.

In plain terms, ask three questions. Will any personal data go into ChatGPT? Is any of it sensitive, about vulnerable people, or used to score or decide things about someone? And is it happening at scale, or as part of a new process?

If the honest answer is “no personal data”, you're unlikely to need a DPIA, but write that decision down. If you're putting in sensitive data or making decisions about people, do the DPIA before you start and involve your data protection adviser. The ICO explains the test in its guidance on when you need a DPIA.

Can you trust what ChatGPT tells you?

Not without checking. ChatGPT can get facts wrong and still sound completely sure, and OpenAI says as much in its own help pages. Treat every answer as a first draft. A person should check anything that goes to a customer, anything with a figure in it, and anything that relies on the law.

OpenAI's help article Does ChatGPT tell the truth? is frank about this. It matters for data protection too, because UK GDPR requires personal data to be accurate. If ChatGPT invents a detail about a customer and it ends up in your records, your business is responsible for that record.

What does Acas say about staff using AI?

Acas recommends clear policies on AI at work, and says employers should consult staff when introducing it. It warns that expecting people to use AI could change their terms and conditions, that outputs need checking for accuracy, tone and bias, and that your data privacy policies still apply. Staff should use approved platforms.

That comes from Acas's May 2025 advice to employers, which also says staff should be careful putting business-sensitive or personal information into public tools. Its research found 26% of workers worry AI will lead to job losses, so how you introduce it matters as much as the tool.

My AI policy template for UK businesses turns this into one page of rules. Where a rule changes someone's contract or feeds into disciplinary action, get advice from your HR adviser or an employment lawyer.

Common pitfalls

• A ban you can't enforce. It risks pushing use onto personal phones, out of sight and outside your rules.
• Relying on Temporary Chat. It keeps chats out of training, but they still go to OpenAI and are kept for up to 30 days.
• Assuming anonymised means anonymous. Removing a name doesn't help if the rest of the detail points to one person.
• Forgetting connected apps. Decide which apps are allowed before anyone clicks connect on SharePoint, Outlook or Google Drive.
• Nobody checking the output. A wrong answer in a customer email is still your business's mistake.

What to do next

1. Find out what's happening now. Ask your team, without blame, which AI tools they use and for what.
2. Move work onto business accounts. ChatGPT Business or Enterprise, or Copilot Chat if you're on Microsoft 365. My Copilot vs ChatGPT comparison helps you choose.
3. Check the settings. Confirm training is off, decide how long chats are kept, and choose which connected apps are allowed.
4. Write one page of rules. What's safe to paste, what never is, and who checks the output.
5. Screen for a DPIA. If personal data will go in, run through the ICO's questions and write down your decision.
6. Train the team, then review. Show people how to get good answers without risky data, and look again in three months.

How I can help

My AI training for small business teams covers all of this. We set up business accounts, agree simple rules on what can be shared, and practise on your own tasks, so people get good results without risky data. If you'd like to know where AI could save time first, start with my AI readiness assessment.

UK GDPR is built into how I set things up, and I'll say so plainly when a question needs a data protection adviser or a lawyer instead.

Questions people ask

Is ChatGPT Plus safe for work?

It's a personal plan, so it belongs to the individual, not your business. Chats can be used for training unless the user switches that off, and there's no data processing agreement with your company. It's fine for general tasks with no personal data. For customer information, use ChatGPT Business or Enterprise.

Can ChatGPT see our emails and files?

Only if someone uploads them or connects an app, such as Outlook, SharePoint or Google Drive. On a Business workspace, an admin can switch those apps off. Decide which apps are allowed before your team starts connecting them.

If I delete a chat, is it gone?

It disappears from your account straight away. OpenAI then schedules it for permanent deletion within 30 days, unless it has to keep it longer for security or legal reasons. On Business and Enterprise, your admins also set how long chats are kept.

Do we need to tell customers we use ChatGPT?

If their personal data goes into it, your privacy notice should cover it. The ICO says privacy information must include who receives personal data, or the types of organisation, and any transfers abroad. Ask your data protection adviser to check your notice.

Is Microsoft Copilot safer than ChatGPT?

On their business plans the promises are similar: no training on your data, a data processing agreement and admin controls. Copilot's difference is that it sits inside your existing Microsoft 365 contract and permissions. My Copilot vs ChatGPT guide compares them side by side.

Does UK GDPR apply if we only paste business information?

UK GDPR covers personal data, so purely business information falls outside it. But that information can still be confidential, under a contract or a duty to a client. The same habit works for both: use a business account, and keep anything confidential out unless you're sure you're allowed to put it in.

Related guides