THOKKI
Guides · 7 min read · 15 November 2026

Is ChatGPT safe for business data? A UK GDPR guide

What happens to what your team types into ChatGPT, Copilot and Gemini, what UK GDPR expects of you, and a short set of rules that lets people use AI without a data breach.

Short answer: it can be. Whether it is depends far less on the AI than on which version your team is using, and what they paste into it.

Right now, in most businesses, someone is pasting a customer complaint into a free AI account to help write the reply. They mean well. They've also just shared a customer's personal data with a supplier you haven't checked. Here's how to make that safe.

Free accounts and business plans are different products

  • Free and personal AI accounts may use what people type to improve the provider's models, unless someone switches that off in the settings. You have no contract with the provider covering your data.
  • Business plans, such as ChatGPT's business tiers, Microsoft 365 Copilot and Gemini in Google Workspace, generally don't train on your data by default and come with a data processing agreement. Check the current terms for the plan you use: they change.
  • Tools connected through an API sit somewhere in between, and depend on how they're set up.

If your team uses AI at work, the first job is getting them onto an account you control.

What UK GDPR expects

UK GDPR doesn't mention AI by name. The usual rules apply to any supplier that handles personal data for you:

  • Have a lawful basis for using the data, and only use what you need for the job.
  • Sign a data processing agreement with the provider, and know where the data is stored. If it leaves the UK, there needs to be a legal mechanism for the transfer, which large providers usually include in their terms.
  • Tell people how you use their data. If AI is part of that, your privacy notice should say so.
  • For anything high-risk, such as decisions about people's jobs, credit or health, do a data protection impact assessment first, and keep a person making the decision.

The Information Commissioner's Office (ICO) publishes guidance on AI and data protection, and it's worth a read before you roll anything out.

Five rules for your team

  1. Use the company account, never a personal one, for anything work-related.
  2. Leave out names, contact details and account numbers unless the tool is approved for them. "A customer in Hull" works as well as their full name.
  3. Never paste in health, financial or HR information about a person.
  4. Check what comes back. AI is confident, not always right.
  5. If you're not sure, ask before you paste.

Put those on one page, share it with the team, and you're ahead of most businesses.

What about large organisations?

The same rules scale up, with more moving parts: an approved list of tools, single sign-on, logs of who used what, and connections to company documents that respect existing permissions. That's the part where the set-up matters more than the tool, and where an AI consultancy audit usually starts.

This guide is general information, not legal advice. If you're handling sensitive data at scale, talk to your data protection officer or a solicitor as well.

Straight answers.

Does ChatGPT keep what I type?

Conversations are stored on the provider's servers for a period, and on free and personal accounts they may be used to improve the models unless that setting is turned off. Business plans set out retention and training in a contract. Check the current terms for the plan you use.

Is it illegal to put customer data into ChatGPT?

Not automatically, but you need a lawful basis, a data processing agreement with the provider, and to use no more personal data than the job needs. Pasting customer details into a free personal account usually fails those tests.

Do we need an AI policy?

If anyone in your business uses AI tools, yes. It can be one page: which tools are approved, what must never be pasted in, and who to ask.

Ready to
dive in?

Book 30 minutes. No deck: bring the problem, and we'll tell you honestly whether AI is the answer.