Lorixen.
EN

Pasting Customer Data Into ChatGPT: A UK GDPR Checklist for Staff

Short answer: Customer data pasted into an AI chatbot is still personal data, so UK GDPR applies to it the same way it applies to your CRM. The practical answer for a small business is a short traffic-light rule: green information (public or fully anonymised) can go in, amber information (names, emails, order details) only goes into an approved business account after removing what is not needed, and red information (health, financial, ID numbers, anything about children) never goes in. Pair that with the right account type, training settings switched off where the vendor offers it, and a list of approved tools. This is general guidance, not legal advice; the ICO publishes detailed material on AI and data protection.

Why this needs a rule at all

A receptionist summarising a complaint, a sales assistant drafting a reply to a long email thread, an office manager cleaning up a spreadsheet of client addresses: all reasonable tasks, and all involve pasting someone else's personal data into a third-party service. Without a rule, each person decides alone, and your business has no idea which tools hold which customer details.

Under UK GDPR your business remains responsible for that data. You need a lawful reason to process it, you must keep it secure, and you need to know which suppliers handle it. A free consumer chatbot account created with a personal email address meets none of those checks in an organised way.

Red, amber and green examples

Give staff concrete examples, because abstract categories do not stick.

Colour Examples Rule
Green Your own website text, a public product description, a made-up example customer, a generic email template Fine in any approved tool
Amber A customer's name and order number, a complaint email, a list of client company names and contacts Only in an approved business account, with unnecessary details removed
Red Health information, bank or card details, National Insurance or passport numbers, passwords, HR and disciplinary notes, data about children, anything under a confidentiality clause Never pasted into an AI chat tool

Special category data (health, ethnicity, religion and so on) carries stricter rules under UK GDPR, which is why it sits firmly in red for a small business without a specific assessment.

Redact before you paste

Most tasks do not need the identifying details at all. To get help rewriting a reply to an angry customer, the tool needs the tone and the problem, not the person's name and phone number.

Teach a two-minute routine:

  1. Copy the text into a blank document first, not directly into the chat.
  2. Replace names with roles: "the customer", "our engineer".
  3. Delete phone numbers, email addresses, postal addresses and account numbers.
  4. Swap exact amounts or dates for rough ones if they are not essential to the task.
  5. Paste the cleaned text, get the result, then put the real details back in your own system.

For spreadsheets, delete identifying columns before uploading or describe the structure ("column A is a date, column B is a postcode area") and ask for a formula rather than handing over the file.

Consumer plan versus business plan

AI vendors usually offer different terms for individual and business customers. Typical differences to look for on the vendor's own pages:

  • Training on your content: business and team plans often exclude your inputs from model training by default, while consumer plans may use them unless switched off.
  • Data processing terms: business plans commonly come with a data processing agreement; consumer plans generally do not.
  • Admin control: a business workspace lets you see who has an account, remove leavers and set retention.
  • Retention settings: some business plans let you control how long chats are kept.

Check the current terms for whichever tool you choose, because they change. OpenAI, for example, explains how it uses data to improve its models and how consumer users can turn this off.

If anything in amber will go into a tool, the business plan with a DPA is the sensible minimum.

Settings to check on every approved tool

For each tool on your approved list, an owner or manager should confirm:

  • Training on chat content is turned off, or excluded by the plan.
  • Chat history retention is set to something your business is comfortable with.
  • Staff log in with work accounts, ideally through single sign-on, not personal Gmail.
  • Sharing links to conversations are disabled or restricted.
  • Connected apps and plugins that can read email or files are reviewed before being enabled.

Take a screenshot of each setting with the date. That is your evidence if anyone asks later.

Keep a list of approved tools

One short table, kept where staff can find it:

  • Tool name and plan (for example "chat assistant, Team plan")
  • What it may be used for (drafting, summarising, formulas)
  • Highest data colour allowed (green only, or green and amber)
  • Who manages the account
  • Date the settings were last checked

Anything not on the list is green-only by default. Add the approved tools to your data protection records and privacy notice where relevant.

The one-page staff note

Here is a structure you can adapt. Keep it to one side of A4 so people actually read it.

  1. Why: customer data in AI tools is still covered by data protection law, and we are responsible for it.
  2. Approved tools: list them, with a line on how to request a new one.
  3. Traffic light: the table above, with two examples from your own business.
  4. Redact first: the five-step routine.
  5. Never: passwords, card details, ID numbers, health or HR information, children's data.
  6. Check the output: AI answers can be wrong; a person reviews anything sent to a customer.
  7. Mistakes: if something red was pasted, tell the named manager the same day so they can assess whether it is a reportable breach.

Ask each person to read it and confirm by email. Revisit it when you add a tool or every six months.

FAQ

Is it illegal to paste a customer's email into ChatGPT?

Not automatically. The question is whether you have a lawful basis, whether the tool and its terms are appropriate, and whether you shared more than needed. An approved business account and redaction make it far easier to justify.

Does turning off training make a consumer account safe for customer data?

It helps, but it does not give you a processing agreement, admin control or a way to remove leavers. For anything amber, use a business plan.

What about AI features inside tools we already use?

Email clients, CRMs and office suites increasingly include AI assistants. Check each vendor's AI terms; they may differ from the main product terms, so treat them as a separate tool on your list.

Do we need to tell customers we use AI tools?

If AI processing of their data is part of how you handle enquiries, your privacy notice should describe it in general terms. Check the ICO's guidance on transparency.

If you would rather build an AI workflow where customer data stays inside systems you control, let us know what you need and we can sketch the options.

Sources

WhatsAppBook a call