It’s one of the first questions people ask us, and we think it’s a very good sign: people who worry about their customers’ data are the ones who’ll look after it best. So let’s go through it calmly and without legal jargon, because the rules are more reasonable than they look.
Who is responsible for what
When a visitor writes in the chat on your website, the data they leave (their questions and, if they choose, their name or phone number) belongs to your business. In the language of the General Data Protection Regulation, the GDPR, your business is the controller. The chatbot provider is the processor: it handles the data only to provide the service to you and on your instructions.
That relationship has to be set out in writing in a data processing agreement. It isn’t an optional formality: it’s what guarantees the provider can’t use your customers’ data for anything else. If a provider doesn’t offer you one, that’s a bad sign.

Where the data is
The GDPR doesn’t forbid data from leaving the European Union, but it requires safeguards when it does. For some years this was a headache with the United States, because the earlier agreements were struck down by the EU’s Court of Justice. Since July 2023 there has been the EU-US Data Privacy Framework, which allows data to be transferred to US companies that have signed up to it with a level of protection the European Commission considers adequate.
Even so, the safest option is for the data to be stored in Europe and, if any service provider processes it outside, for that to happen with the safeguards the GDPR requires and to be explained in the privacy policy.
The question almost nobody asks: do you train your AI on my data?
This one matters. Some free or consumer AI services may use what people type into them to improve their models. For a business that’s a problem: you don’t want conversations with your customers ending up as part of something you don’t control.
Ask directly and get the answer in writing. The right answer is a flat no.
Telling people they’re talking to an AI
The EU AI Act adds a simple, sensible obligation: people have to know they’re talking to an AI. A chatbot that pretends to be a person is dishonest, and it also breaks the rules. It should introduce itself as your business’s AI assistant and, if someone wants to talk to a person, let them.
Consent when people leave their details
If the chatbot asks for a name, email or phone number, it has to do it properly: explaining what they’ll be used for, with a consent tick box and a link to your privacy policy. And visitors must be able to exercise their rights (access, rectification, erasure) as with any other data you hold about them.
Cookies
Another detail that sometimes gets overlooked: some chat tools add cookies to your website, which means you have to ask for consent in the cookie banner. If the chat doesn’t set cookies, you avoid that problem.
What to ask your provider
- Will you sign a data processing agreement with me?
- Where is the data stored?
- Do you use the conversations to train any AI?
- Does the chat introduce itself as an AI assistant?
- Does the contact form collect consent?
- Does the chat set cookies on my website?
Our commitment
At OX·IA we take all this very seriously, because we know it’s what allows a business to trust a tool like this. The data processing agreement is included in our terms, the servers are in Spain, your customers’ data isn’t used to train any AI, the chat doesn’t set cookies on your website, the assistant always introduces itself as your business’s AI and the contact form collects consent with a link to your privacy policy. If any service provider processes data outside the EU, it does so with the safeguards of the GDPR, as explained in our privacy policy.
For us, data protection belongs in the product itself, not in a legal section at the bottom of the website.



