Does AI train on your business data?

On the business plans from Anthropic (Claude) and OpenAI (ChatGPT), your data isn't used to train their AI by default, but on personal plans - including the paid ones - it depends on a privacy setting that many people have never opened.

That's the short version. The detail matters, because this is a subject where a lot of what's online is unreliable. While checking the facts for this article we found several sites making contradictory claims about the same provider's policy. So everything below comes from the providers' own documentation, with links, and it's worth clicking them rather than taking our word for it.

What does "training on your data" actually mean?

Think of an AI model as a student who learns by reading a huge pile of past papers. "Training" is the reading. If a provider trains on your conversations, what you typed can become part of that pile, helping to shape how future versions of the model answer everyone.

It doesn't mean someone is reading your chats to find out about your business. But if what you pasted in included customer details, pricing or anything confidential, most businesses would rather it stayed out of the pile altogether.

Does paying for a plan stop it?

Not on its own. Most people assume the dividing line is free versus paid - that paying makes you the customer rather than the product.

The line that actually matters is personal versus business. Claude Pro and ChatGPT Plus are personal, consumer products. Upgrading from free to one of those, or to a more expensive personal tier, moves you up the consumer ladder. It doesn't move you across to the business side. What does that is a team, business or enterprise plan, or using the API.

The API is worth a quick explanation. It's a back door into the AI that other software is allowed to knock on, so your own systems can send it work directly rather than a person typing into a chat window. Business tools that "use ChatGPT" or "use Claude" behind the scenes usually go through it.

What Anthropic says

On business plans: "By default, we will not use your inputs or outputs from our commercial products (e.g. Claude for Work, Anthropic API, Claude Gov, etc.) to train our models." That's from Anthropic's own privacy documentation. "By default" is the important part - training is off unless you turn it on.

On personal plans (Free, Pro, Max): training is a setting you chose, or had chosen for you by not answering. Under the consumer terms updated in 2025, if you allow your data to be used for model improvement, it's kept for five years. If you decline, it's kept for thirty days. Existing users were asked to choose by 8 October 2025. If you clicked through that message without reading it, as many people do, you may not know which way you went.

What OpenAI says

On business products: data from ChatGPT Business, Enterprise, Edu and the API Platform isn't used to train models unless you explicitly opt in. API data has worked this way since 1 March 2023.

On personal plans (Free, Go, Plus, Pro): data may be used for training, depending on whether you've opted out.

Two different companies, the same shape. On business plans the default is off. On personal plans it's a setting, and the setting isn't obvious.

Two things worth doing this week

Look at your setting. Not your plan - your setting. In both products it's in the privacy settings and takes under a minute to check. Changing it only affects what happens from now on. It doesn't pull back conversations that have already been used, although both providers exclude deleted chats and closed accounts from future training.

Work out whose account your business data is going into. For most small businesses this is the bigger question, and it's the one fewer people think about.

Your staff may already be using their own accounts

The Office for National Statistics published a figure in its survey of AI in UK businesses that deserves more attention:

55% of employees report using AI personally. Only 35% of businesses report using AI at all.

That gap is people using their own accounts to do their job - pasting a customer email into a free ChatGPT tab to get a better reply written, running a spreadsheet of figures through it, drafting a quote. It's usually done with the best of intentions, but it happens on personal accounts with personal-plan defaults, where you can't see it. There's more on that gap in our article on how many UK businesses actually use AI.

For a small business, this is the real issue. You can be careful with your own settings and still have your customer list, your pricing and your contract wording sitting in accounts you've never seen.

UK GDPR doesn't care whose subscription it was. If a pasted email contained a customer's personal data, your business is still the "controller" of it - the one responsible for looking after it, in the same way you'd be responsible for a customer's file left on a train by a member of staff. The rules you already follow still apply, and a new tool doesn't come with an exemption. The ICO has been clear that organisations "developing or using generative AI must consider data protection from the outset", in its statement closing its investigation into Snap's My AI chatbot.

What we'd suggest instead of banning it

Banning AI tends to leave you where you were, except nobody tells you when they use it.

The more useful approach is quieter. Decide which account the work happens in, and pay for the kind of plan where training is off by default. Say plainly what may go in and what may not - customer personal data, anything under a non-disclosure agreement, anything you wouldn't want read aloud. Write that down once, on a single page, rather than mentioning it at a meeting.

It's the same conversation businesses once had about file-sharing services and personal phones, and it tends to end in the same place: give people a proper, approved way to do the thing they're going to do anyway.

If you'd rather not work out which business AI plan fits and get it set up for everyone, we're happy to do that for you.

The good news is that the approved way is genuinely better, not just safer. A business plan isn't simply a licence with the training switch turned off. It's where the account controls, the shared information and the connections into your actual systems live. That's the point where AI stops being a chat window someone pastes into and starts doing useful work on your own data. If that's the part you're interested in, you can see what we set up for small businesses, and we're happy to help connect AI to the tools you already use.

Sources

• Is my data used for model training? - Anthropic Privacy Center
• Updates to our Consumer Terms and Privacy Policy - Anthropic
• How your data is used to improve model performance - OpenAI
• Artificial intelligence in UK businesses: 2023 to 2026 - ONS
• ICO warns organisations must not ignore data protection risks as it concludes Snap 'My AI' chatbot investigation - ICO

Checked against the above on 8 September 2026; the ICO quote was corrected and checked on 23 September 2026. These policies change more often than most, so check the links rather than trusting the date on this page.