Guide
AI and UK GDPR: a plain-English guide for small businesses
What a small UK business should check before using AI tools with customer data: tool terms, data processing agreements, telling people, and human review.
By Rawleaf · · 9 minute read
AI tools are now built into the email, office and accounting software many small businesses already use, and lots of people in your team are probably trying general-purpose AI assistants on their own. That's often a good thing. But the moment customer, patient or staff details go into those tools, UK data protection law applies, just as it does to the rest of your business.
None of this needs to be daunting. Most of it is common sense, written down. This guide covers the practical questions a business owner or manager should be able to answer.
This is general information, not legal advice. The law and the official guidance change, and every business is different. For the legal detail, the ICO's guidance on AI and data protection and its hub for small organisations are the places to start. If you're unsure about something specific, take professional advice.
The short version
- Don't put personal data into consumer AI tools. Use business versions set up properly.
- Check what a tool's terms say about how your data is used, stored and kept.
- Make sure there's a data processing agreement in place with any AI provider that handles personal data for you.
- Tell your customers and staff, in your privacy notice, how you use AI with their information.
- Keep a person in charge of decisions that affect people.
- Take extra care with sensitive information, such as health details.
What counts as personal data
Personal data is any information about an identifiable living person. In everyday business that includes names, email addresses, phone numbers, home addresses, photos of people or their homes, messages they've sent you, and notes about them.
Some kinds of information need extra care. Health information, for example, is "special category" data under UK GDPR. If you run a clinic, or if customers ever tell you about their health, this applies to you. The ICO explains the extra conditions that apply.
A useful habit: before pasting anything into an AI tool, ask "could this identify someone?" If yes, stop and check the tool is set up for it.
Consumer tools and business tools are not the same
Many AI assistants come in a free or personal version and a business version. They can look identical, but the terms behind them often differ in ways that matter:
- Whether your data is used to improve the provider's models. Personal versions may allow this by default. Business versions often don't, but check.
- Who is responsible for what. Business versions usually come with a data processing agreement. Personal versions usually don't.
- Admin control. With a business account you can see who is using it, manage access and close accounts when people leave.
The simplest rule for your team: no customer, patient or staff details in personal AI accounts. Anonymised examples for testing ideas are fine. Real people's details are not.
What to check in a tool's terms
You don't need to read every word, but you should be able to answer these questions for any AI tool that will handle personal data. The answers are usually in the provider's terms, privacy policy or data processing agreement.
- Is our data used to train or improve their models? You want a clear "no" for business use.
- Where is the data stored and processed? If it leaves the UK, there are rules about international transfers. Reputable providers explain how they meet them.
- How long is it kept? Including chat history, uploaded files and logs.
- Who else handles it? Providers often use other companies (sub-processors). They should list them.
- How is it secured? Look for clear information on encryption, access controls and how they handle security incidents.
- Is there a data processing agreement? See the next section.
If you can't find the answers, ask the provider. If they can't answer, that tells you something.
Data processing agreements, in plain English
When an AI provider handles personal data on your behalf, your business is usually the controller (you decide why and how the data is used) and the provider is a processor (they handle it on your instructions). UK GDPR requires a written contract between you, often called a data processing agreement or DPA.
For most business-grade AI services, this is part of the standard terms, and you accept it when you sign up or switch it on in your account settings. What matters is that:
- it exists, and applies to the account your team actually uses;
- you've kept a copy, or know where to find it; and
- it says the provider only uses your data to provide the service to you.
The same applies to anyone who builds AI tools for you. If they'll handle your customers' data, there should be an agreement covering it.
Telling your customers
People have a right to know how their information is used. In practice that usually means updating your privacy notice to say:
- that you use AI tools to help with things like replying to enquiries, drafting documents or organising information;
- what kind of information goes into them;
- who the providers are, or what type of providers they are, and where data is processed; and
- how people can contact you with questions or to use their rights.
If customers talk directly to an AI assistant, for example in a chat on your website, it's good practice to tell them it's an assistant, not a person, and how to reach a person.
Keep a person in the loop for decisions about people
AI is good at drafting, sorting and summarising. It shouldn't be the one deciding things that significantly affect people, such as whether someone gets a job, a tenancy, a service or credit.
UK GDPR has specific rules about decisions made solely by automated means that have legal or similarly significant effects on people, and the law in this area has been changing. Check the ICO's current guidance if you're considering anything like this.
The safer and simpler approach for most small businesses is the one we use anyway:
- routine replies in wording you've approved can go out automatically;
- anything that commits the business, or makes a decision about a person, waits for a person to review it and decide; and
- that review is real, not a rubber stamp. The person has the information and authority to change the outcome.
When to think harder
Some uses carry more risk and deserve more care before you start. For example:
- anything involving health information or other special category data;
- using AI to assess, score or monitor people, including your own staff;
- processing large amounts of personal data, or combining data from different sources.
For higher-risk uses, the law may require a data protection impact assessment (DPIA). This is a structured way of thinking through the risks and how you'll reduce them, done before you start. The ICO explains when one is needed and provides a template.
A short checklist for your team
If you do nothing else, agree these as a team and write them down:
- Which AI tools are approved for work, and which account to use.
- No personal data in personal or free AI accounts.
- Anything AI drafts that goes to a customer is checked, unless it's a routine reply you've approved in advance.
- Decisions about people are made by people.
- Who to ask if you're unsure.
A one-page AI policy covering these points is a sensible start, and is more useful than a long document nobody reads.
How we can help
Every AI Health Check includes plain-English UK GDPR and risk notes for the specific ways your business might use AI, so you know what to put in place before anything is built. If you're running several AI tools or suppliers and want someone senior to keep them on track, including how they handle your data, that's part of our Part-time AI Lead.
Either way, the ICO remains the authority on the law, and for anything specific to your situation you should take professional advice.
More for your industry: AI for tradespeople, AI for letting agents, AI for clinics, AI for accountants