How to use AI in your marketing without breaking GDPR
To view this video, please accept marketing cookies.
Posted: Fri 20th Nov 2026
AI tools make marketing faster. But pasting customer details into the wrong one, or connecting it to the wrong system, can put you on the wrong side of GDPR.
Most small businesses now use AI somewhere in their marketing, often through free accounts and without checking where the data goes.
In this Lunch and Learn, Anna Morrish, founder of multi-award-winning agency Quibble, walks through the practical checks to make before you use an AI tool with business or customer information.
It's a plain-English session for business owners and marketers, not legal advice. Leave with a simple checklist you can use straight away.
Topics covered in this session
What you should and shouldn't paste into AI tools, and why free consumer accounts carry more risk
How to check whether a tool trains on your data, and where to switch that off
What to check before connecting AI to your inbox, CRM or Google Drive, and why a tool should only see what it needs
About the speaker
Anna is the founder and managing director of Quibble, a multi-award-winning digital marketing agency. She has spent more than a decade in SEO and digital marketing, and today her work centres on marketing and AI.
She helps businesses understand how tools like ChatGPT and Google's AI Overviews choose which brands to recommend, and how to make sure theirs is one of them. She also helps them use AI automation to save time without losing the human touch.
Anna speaks regularly on marketing and AI, including three times at the world's biggest SEO conference and at Enterprise Nation events. Her talks are practical and jargon-free, so business owners leave with steps they can act on straight away.
Watch more expert webinars
Access a growing collection of expert-led webinars covering marketing, sales, finance, growth and more – ready whenever you are.
Transcript
Lightly edited for clarity.
Caitriona O’Connor: Hello, everyone, and welcome to today’s Lunch and Learn. My name is Caitriona, and I’ll be your host today. For those of you attending a Lunch and Learn for the first time, Enterprise Nation is a vibrant community platform for start-ups and small businesses.
I’m pleased to introduce Anna Morrish, founder of Quibble. In this session, Anna will talk you through the practical checks to make before you use an AI tool with business or customer information.
If you have any questions throughout the webinar, please post them in the chat, and we’ll do our best to answer them at the end of the session. Today’s webinar will be recorded, and we’ll send you a follow-up email with the recording and further resources later today. Over to you, Anna.
Anna Morrish: Brilliant, thank you. Thank you very much for joining us this afternoon – hopefully, I’m going to give you lots of tips and good advice.
We’re here to talk about how to use AI in your marketing without breaking GDPR. As usual, there’s a theme – if you’ve been to any of my talks or seen my content, you’ll be familiar with that – and it’s there to lighten the mood whilst we talk about this particular topic.
So first of all, welcome aboard, my scallywags. You may be wondering why we’ve gone with a pirate theme alongside GDPR. Well, that’s because a lot of us are using AI a bit like pirates.
We grab whatever’s useful, we paste in whatever’s at hand, and we assume no one’s really watching. Unfortunately, in every pirate film, the navy turns up eventually – or the ICO, in this case.
I’m your guide today – your captain. But to be clear, I’m a marketer, not a solicitor. I’m the founder of Quibble, a digital marketing agency that’s been going for just over 10 years.
I spend most of my days helping small and medium-sized businesses get found on Google and recommended in AI, so I see a lot of this stuff. But first of all, I need to say that AI isn’t the problem – the careless use of AI is.
The team here at Quibble and I use AI every single day. We’ve seen what happens when it’s used without any rules, and we hear horror stories from our clients as well.
I’m not here to tell you to stop using AI. And as I say, I’m not a solicitor, so if you’re doing something big or risky with personal data, I’d recommend getting proper advice. Hopefully, you won’t get yourself into that situation.
Whilst you’re here, tell us where you’re joining from. Who are you, and what business are you in? Let everybody know in the chat and get the conversation going.
We’ll have a few Q&As as well, so I’ll be asking you to pop some details and opinions in the chat. And be honest – have you ever pasted a customer’s email or details into ChatGPT, Gemini or another AI tool?
There’s no judgement. It’s happened, and a lot of people don’t have the guidance – that’s what I’m here today to give you as your captain.
So where are we going on today’s voyage? We’ll stop at a few different places: what counts as treasure, what the ICO says, what breaks GDPR and what free AI accounts get you. We’ll also talk about training, connecting the tools you’re currently using, and what you should and shouldn’t share.
And of course, we’ll have a checklist at the end – that’s where the gold is. But not all treasure is silver and gold.
In business, the treasure isn’t just the money. It’s your customer list, your enquiries and the data coming through – it’s staff records, it’s your pricing. The issue is that when we paste it into AI, we’re essentially handing over the map to our very own treasure, and that’s something we don’t want to do.
But is using AI stealing? I’ve included this social post because it says AI is built on stolen data, that it’s poisoning the planet and various other things. This post in particular got 130,000 reactions in just two days.
Lots of people think that, yes, AI is built on stolen data. You don’t have to agree, and I’m not going to argue either way. But it matters because your customers are the ones liking and engaging with this type of content.
So how their data is used and processed is very important, and being open about how you use it is very important too. Here’s a question for you: do you think using AI is stealing? Type yes or no in the chat – I’d be interested to see what you think.
Now, an interesting stat – and this is the scary part. 71% of UK employees have used AI tools that their employers haven’t actually approved.
And only 32% of people actually worry about the customer data they’re putting in. By being here, you’re taking the first step to mitigate these issues – and you clearly do worry about the customer data you’re putting in.
46% – so nearly half of workers – are using free AI tools, and 31% are using them without their employer even knowing. We’re all scrambling to keep up with what’s happening in AI. The pressure is on, and that’s pushing people to use these tools a bit willy-nilly.
Why does it matter? The big reason is that free accounts are the ones most likely to learn from what you put into that AI bot – more on that at stop four.
Moving on to personal data, what actually counts as personal data? Basically, anything that identifies a living person. That could be names, emails, phone numbers and photos – photos are a big one.
Before this session, I was saying that a lot of people are now receiving their children’s photos from schools. Many of those photos have the school emblem on them, or a file with the child’s name. So when you let your child be in that position, it’s important to understand whether the school is using AI, whether they’re editing the photos in any way, and what their process and policy are.
It’s quite important, because these are our children. You may have already seen posts about it, as it’s been a bit of a hot topic, especially if you’ve got kids in school and are in various Facebook groups.
But that’s not the only thing. Personal data also includes IP addresses, order history, and voice notes that may contain information identifying a particular person. I won’t read all of this out – we’ll be sharing the slides afterwards.
Another thing I’ll mention is that company secrets aren’t covered by GDPR. If you’re adding details about your services, that’s absolutely fine. But some things don’t necessarily belong in a free chatbot – your company financials, for example.
So what does GDPR actually ask of you? People have talked about GDPR till the cows come home – it’s been a hot topic for many, many years. Basically, it’s about how you handle people’s data: your customers’ and your staff’s.
It’s really important to be fair and upfront about how you’re collecting that data and how it’s used. Even with people internally, make sure they understand how their data is being processed. And stick to the reason you collected it.
I mentioned in a previous talk with Enterprise Nation about AI slop that we use AI for business prospecting. The data I’m collecting is all public information – usually contact@ or support@ email addresses, where I’m asking for my message to be forwarded on. But even so, I’m not going to take that data and put it into a mailing list or use it to build out an ad campaign.
I’m only using it for a specific reason. If they don’t respond, they don’t respond, and that data doesn’t go anywhere else. And if any personal information comes through, we have a policy in place that says we’ll remove it from our system after a set number of months or years – that’s important too.
So those are the seven common-sense rules. It’s important to note that none of them mentions AI, and that’s the point. AI doesn’t change these rules – it just makes them easier and faster to break.
Using someone else’s AI tool doesn’t make them responsible. This is the bit people often miss. You might think, “It’s OpenAI, it’s ChatGPT – it’s their problem if they learn from the data I put in.”
But no, the ICO says it’s not. You can’t avoid it – it’s your responsibility, and it’s your problem if you break the rules. This is what they say, directly from the guidance on their website: accountability lies with the controller – your organisation.
If you need to demonstrate how you’ve used AI and how the system complies with data protection legislation, you can’t just say you told someone how to use it. You need to make sure there are accountable processes in place, and in most cases, you’re the one who’ll be responsible.
You’re the responsible person for data protection – essentially, you’re the data controller. You’re the one who decides why and how the data you hold is used. So it’s very important that you’re up to speed with best practice and that it’s followed from the top down.
As the business owner or founder, you have to follow these rules yourself and then make sure your team understands them too. By being here today, you’re making a start.
This is also information straight from the regulator. When I was putting these slides together, I saw they’re currently running a survey – it closes today – asking people how they use AI in their business, what’s important to them and what they’re concerned about. They’re doing this because they’re trying to provide more guidance to businesses, as there isn’t a lot yet beyond the standard GDPR material.
This information is all live on the ICO website, and I’ll share some links on my LinkedIn after this. It’s also in the slides we’ll share after the talk. I’d recommend going to the ICO website and reading what they have on artificial intelligence.
But be sure you don’t hand this off to anyone else – it’s got to be with you. You’re the responsible person who has to be up to speed with all this information. You can’t delegate it to your team, whether that’s the tech team or anyone else – it’s got to stick with you.
There are various reasons why, because this is what getting caught looks like. This Samsung story is from 2023, and obviously we’ve come a long way since then. At the time, they banned the use of AI within the organisation, though I’m sure that’s no longer the case.
In that instance, their staff were just trying to be helpful. They weren’t thinking, “We’re going to paste company data into a free AI tool and be malicious.” That wasn’t their intention, but it did lead to a code leak, and there were fines and other consequences involved.
Another company, Replika, was fined €5 million. Then there’s Air Canada, whose chatbot gave a customer the wrong information and advice. The responsibility lay with Air Canada, not the chatbot.
Even if the chatbot made a mistake, it’s not a technical issue – it’s the company’s issue, because they will come for you. What I find quite staggering is that fines are increasing significantly.
Marketing emails and texts, for example, are covered by something called PECR, which sits alongside GDPR. Until this year, the maximum fine was about £500,000 – still quite significant. But since the Data (Use and Access) Act came in, it’s increased to £17.5 million or 4% of your turnover.
So if AI helps you send more emails – as it does for us, for example – check that you’re allowed to send them. And don’t automatically opt people into lots of email marketing lists.
You’re probably thinking, “But I’ve been opted into so many marketing lists, and I’m getting floods of emails I never signed up to.” That’s because people are doing it incorrectly, but that doesn’t mean you should be one of them. It only takes one person or a handful of people to report those emails to the ICO for them to investigate and for you to potentially have an issue.
So, walking the plank – the things that are likely to sink you and get you into trouble. You might already be doing some of these things, and that’s what we’re going to discover now.
First off, what isn’t GDPR compliant? We’ve looked at what GDPR asks of you, and again, it’s important to note that these things aren’t evil – they’re things people do in a hurry. That’s why it’s a problem: people are rushing, and they haven’t got the guidance they need.
Things like pasting in customer lists, emails and CVs. CVs are an important one – people might get floods of them for a job they’ve recently advertised. They add all that detail and ask AI to pick out the top five candidates to invite to interview, and that goes against GDPR.
Those CVs will have people’s names, likely their address or where they’re from, their email address and maybe their phone number. So don’t put that into an AI tool, especially a free account – that’s very important.
Then there’s using AI to profile or target people in ways they wouldn’t expect. I mentioned that we do business prospecting, but I’m not going to add those people to an ad campaign and start harassing them with loads of ads.
Connecting a new AI tool to customer data with no processing agreement in place is another one. If you haven’t got anything on your website or in your contracts, be sure to add information about your AI usage and how data is processed.
And I think this one’s important: letting AI make decisions about people with no human check. If you listened to my last talk on AI slop, it was all about content, but everything you output with AI should be checked by a human – and that includes the input.
If you’ve got CVs, it’s absolutely fine to get them checked, as long as you black out or delete the information that’s personal to that person. That way, you’re just including their experience and perhaps some details about their hobbies and who they are as a person, but nothing that identifies them specifically. So check the input and the output before you send anything out – that’s very, very important.
Now for a bit of interaction: parley or plank. I want you to vote as we go through these one at a time. For each one, type P if you think it’s fine to put into an AI chatbot – that’s parley – or W if you think you’ll walk the plank and get into trouble.
I’ll give you a few seconds for each, and you can write one to five with a P or a W for each. First, the rota: pasting staff names and shifts into a free ChatGPT account to build out a rota. Is that P, it’s fine, or W, you’re walking the plank?
Number two is the blog: asking AI to draft a blog about all of your services. Again, P if you think it’s fine, W if you think it’s walk the plank.
Number three, the complaint – this is a good one. Pasting in an angry customer’s email, with their name and order number, to draft a reply. You may have done this yourself, or someone may have done it to you when you’ve complained about something.
Number four is reviews: summarising reviews with the names swapped for Customer A, B and C. Consider some of the nuances with these, because some of the information may still need reviewing by a human.
And number five, the CVs: uploading 30 CVs to a free tool and picking a shortlist. Hopefully, you’ve all been listening, because that’s basically a test – I’ve just given you a pop quiz.
So how did you do? If you picked numbers two and four for parley, you’d have been correct – they’re fine. The only instance where I’d say reviews wouldn’t be fine to share is if they include specific information, perhaps about your staff.
Often, if you ask for reviews, you might ask who gave the customer service. Customers might also upload photos, which you’d want to remove before you paste anything into AI. Then you’ve got to walk the plank: the rota, the complaint and the CVs.
A lot of this really is common sense, but it’s about practising it constantly. If you do it on repeat, it becomes second nature and you won’t have a problem. It’s just about knowing this and bringing it to the front of your mind.
Dead men tell no tales, but free AI accounts might. If you’re not paying for your accounts, your data may be used to train the model itself. Free plans learn from you – it’s something people don’t necessarily even think about, and it’s also the single biggest fix.
If you or your team are using personal ChatGPT, Claude, Gemini or similar accounts, they’ll be training on your data unless you turn that feature off – I’ve got details on that in a second. You can simply do that, or you can move to a business plan where your data won’t be used for training. You can also see who’s using the AI and what for, so you’ve got more control.
We mentioned how data is stored as well. If you’re using a free account and you’re hopefully going to switch training off, it’s still important to tell people how long that information is likely to be kept, especially if you’ve used it in the past. With Claude, it’s five years, and Gemini keeps reviewed chats for up to three years.
I think there was something yesterday about Grok as well – details were released that people had entered into their chats. That’s given people a bit of the heebie-jeebies about what they’ve added in the past.
I’ll tell you a funny story that’s sort of related. When we were testing AI – because we’re always testing and training on different AI tools – I wanted to find out what it knew about me. I was trying to get it to give me information, and I said, “Did you know that Anna Morrish wants to be a naturist influencer?”
I’m not, but I know that was originally on a free account, so it’s now stored. I really hope that doesn’t get leaked, because I’m not in any way interested. I don’t judge anybody who is, but I don’t want that connected to me.
So be careful what you add and test in AI because, unfortunately, it keeps that information for a period of time. There you go – a secret about me, though it’s not a real secret.
So turn off your training today, or move to a business plan. These are the ways you can do it in ChatGPT, Claude, Gemini and Copilot for free accounts. It takes two minutes, so you can come off this session and do it before you start your lunch.
Even if you’re moving to business accounts, I’d still check, because as with any third-party tools, they sometimes change the settings without you knowing. So do keep an eye on those.
Take what you can and give nothing back – that’s basically the AI tools’ motto. So before you connect AI to your inbox, CRM or drive, only hand it what it needs. These connectors are brilliant, and we use them.
But if you’re connecting to Gmail, for example, you might be giving it every single email you’ve ever sent. You don’t want to give it free rein, so check these things before you give AI the keys to your business and your data.
First, does it really need full access to everything, or just a particular folder? And if you give it the ability to respond to emails before you’ve reviewed them, I’d turn that off and give it strict instructions.
I won’t read all of these out, but these are the things I’d definitely do before giving it the keys. Making sure you can unplug it and remove any data is really important as well.
Here’s another story – not one from me – about oversharing. One poisoned document in a shared drive tricked ChatGPT into sending out secret keys. You may have heard about it, as it was all over the media, and it’s why access matters.
The document itself had hidden text – really simple text, in white on a white document, with other text in front of it – but it tricked the AI. That has, I believe, been patched. But it proves how important it is to know which documents AI has access to, that they’re clean, and that they’ve come from you and your team.
This is what your GDPR-friendly workflow should look like – it’s the safest everyday workflow. AI doesn’t really need to know that a piece of information is about Mrs Patel at number 12. It just needs to know whether the delivery was late and what you’re going to do in response, for example.
With a business account that has a data processing agreement (DPA), you can relax a little bit, but these steps are never really wrong. So take out the names and details, and scramble the information a little. Check everything, including the facts and the tone, and make sure it’s all coming from your own systems.
Even pirates had to be upfront about what they wanted. Tell your customers how you use AI and how you use their data, and be open about it.
It’s the same for GDPR: you need to tell people how you’re processing their information, so if you’re using AI tools, include those details as well. People have a right to know what you’re doing with their data, including when AI is involved. And it’s not just your customers – I’m sure you’d want to know how your own data is being processed.
There’s a stat that 63% of UK adults would never trust AI to handle their sensitive personal data, which is quite staggering – that’s well over half. So if you’re handling data with AI tools, be upfront about it and have those systems and processes in place.
Tell people what you’re doing – say it out loud. Tell them when it’s a bot, because AI-personalised emails still need consent or a soft opt-in, and you’ve always got to make sure there’s a human involved. These are four places where you should be upfront.
Here’s another very short story. Earlier, I called to book a table at Zizzi for the weekend, and it was an AI bot – but it was clear as soon as I called: “You’re speaking to an AI assistant.” Not that I liked it – I still prefer a human on the end of the phone – but it was clear I was talking to AI, so be open about it.
It’s really important to remember that 71% stat from earlier: it’s not a people problem, it’s a guidance problem. People haven’t been given information about how to use AI and how not to. If there are no rules, write the rules down yourself.
Remember the Samsung case, where staff weren’t malicious? They just didn’t have the rules. Even as a small business, it’s important to have rules in place, because that’s where most fall down – not because you’re careless, but because nobody’s written anything down.
As I mentioned earlier, if the ICO ever asks and you say, “We told people verbally,” it’s not going to cut it – you’ve got to be accountable. So write the rules, approve them, train your crew and tweak them as you go.
Things are going to change. The tools you use will change, and the way you process data is likely to change too, so make sure you’re keeping your rules up to date – there’s no escaping this.
And here’s your treasure. If you only remember one slide, make sure it’s this one – screenshot it now, and I’ll also be sharing it in the slides. But remember, you’re the captain, so processing that data and informing everybody is on you.
Thank you. It’s a lot to cover in just half an hour, but we’ve got time for a couple of questions. If you have any, pop them in the chat, and hopefully we can answer a couple now.
Caitriona: Thank you so much. If anyone has any questions, please do pop them into the chat.
I can see we’ve had a question from Michael, who’s asking: why is the criteria a free account for some elements as opposed to a paid subscription? I think it means across different AI tools, but Michael, if you want to add a bit more detail in the chat, please do.
Anna: If we’re asking why they train on free accounts, I think it’s just because, in a lot of instances, they can. They’re trying to gather as much information as possible. There have been various AI trends on social media where people paste in their photos and it’s been training on them, but it’s classed as a trend.
So I’d say it’s just because they can. It’s the same with social media settings – they change all the time, and the platforms switch on whatever best benefits them.
Their game is getting as much information out of their users as possible, so of course they want to train on what people are doing right now. If there’s anything more specific you wanted to know, feel free to reach out to me on Enterprise Nation or LinkedIn.
Caitriona: Perfect. I think that’s been answered.
Anna: Again, if you have any follow-up questions, or want to be pointed in the right direction on the ICO side, feel free to reach out – I’m happy to help. Are there any other questions, or is that it for today?
Caitriona: That might be it for today – we’ve come to the end of the session. I’m just posting Anna’s links in the chat again, and we’ll also share them in the follow-up email, so please do feel free to reach out, connect and ask any questions you might have.
We’ll also post a link for feedback in the chat. Thank you so much, everyone, and thank you, Anna, for your presentation. Thanks again for joining, and I hope everyone has a great weekend.
Anna: You’re welcome. Thank you.
Explore our webinar library
Unlock more on-demand sessions designed to help you sharpen your skills, grow your business and stay one step ahead. Find more Lunch and Learn webinars
I help small businesses use AI to get found online and to cut down on the repetitive marketing work that eats into their week.
I founded Quibble in 2016 after working as a magazine editor, content marketer and outreach manager, both in-house and agency-side. Today we're a small team in Rutland. We support businesses in agriculture, manufacturing, retail, automotive and tech with SEO, content, paid media, digital PR and web development, and some clients have been with us for more than seven years.
More and more of my work now focuses on AI: how businesses show up in tools like ChatGPT and Google's AI Overviews, and how to use AI to automate everyday marketing tasks like content briefs, reporting and follow-up emails. I've spoken three times at the world's biggest SEO conference, and I'm currently building an AI search training programme for SMEs.
Here on Enterprise Nation I share practical, jargon-free advice you can use the same day. Book a free session if you'd like a sense-check on your marketing or want to know where AI could save you time.