re:solved, AI automation agency
← Back to blog
Cătălin Andrei Bora, founder of re:solved

By · Founder, re:solved

·8 min read

Is it safe to give your customer data to an AI? GDPR and automation, explained

Where customer data goes, what GDPR requires and why local hosting does not automatically mean entirely local processing.

A locked filing cabinet next to a softly glowing server rack in a small European office at dusk.

You type a customer's name into an AI tool and a small voice at the back of your head asks "where did that just go?" That instinct isn't paranoid. It's the right question. The one most vendors quietly hope you won't ask. This piece takes it seriously.

We'll keep it calm and factual. By the end you should feel more informed and less anxious. Not scared into buying something. Fear-marketing is what everyone else does; we'd rather tell you how things actually work.


What actually happens to your data in a typical cloud AI tool

When you type something into a standard cloud AI tool, that text leaves your computer and lands on the vendor's servers. Often outside the UK and the EU. It's processed there under their terms of service, and depending on settings you may never have opened, the data may be retained for a while, or even used to improve future systems.

None of this is sinister on its own. But it means something simple: you've made someone else responsible for data your customers gave to you. Under GDPR, you stay accountable for it. No matter how many vendors handled it on the way.


What GDPR actually requires of your company (the 4-point version)

No lawyer paragraphs. Four things:

  • You stay responsible for customer data even when a tool processes it for you.
  • You need a lawful basis for the processing. The same as with any other software you already run.
  • You must know where the data goes, and have a contract with whoever processes it.
  • Your customers keep their rights (access, deletion) no matter which tool touched their data.

The reassurance: none of this bans AI. What it bans is not knowing what your tools do with the data. The difference matters.


The deadlines. Get them right (this is where everyone else is wrong)

A lot of what you'll read online muddles the timelines. The reality is simpler:

  • GDPR already applies. It has since 2018. Cumulative fines run into the billions of euros. That's the real, present-tense obligation.
  • The EU AI Act adds a transparency duty that has applied since 2 August 2026: if a system talks to people, they must be told it isn't human. The heavier high-risk obligations, which include tools used in recruitment and hiring, apply from December 2027, after the EU deferred the original August 2026 date. That's time to get ahead of it calmly, not an emergency.
  • For UK readers: the UK is not under the EU AI Act (unless you serve customers in the EU) and runs UK GDPR, which for this article's purposes asks for the same things.

One technical point, in plain language: the EU's own data-protection authority has warned that AI systems rarely truly anonymise personal data. "We removed the name" is not the same as "this data is no longer personal."


The self-hosted answer (where we arrive, not where we start)

Live agents run in isolated instances on re:solved’s server in Timișoara. Message text is sent to the OpenAI API under a data-processing agreement. Entirely local hosting and processing are design options, not descriptions of live deployments. Local hosting alone does not guarantee compliance.

Local processing can matter for clinics, accountants or lawyers. We do not have a law-firm client confirming this scenario.


The honest part: self-hosting isn't a magic compliance wand

To be clear: running the agent on your own hardware doesn't remove GDPR from your plate. You still need a lawful basis. You still answer to your customers. And an agent should never be given data you wouldn't hand to a new employee on day one.

If any provider tells you their product makes GDPR "not your problem", that's the moment to walk away. Nobody can take legal responsibility away from you with a contract. They can help you manage it. They can't dissolve it.

GDPR isn't the only compliance rule already in force for AI assistants: What the EU AI Act actually asks of a business using an AI assistant


Five questions to ask any AI provider (including us)

  1. Where does my data physically live?
  2. Is it used to train or improve anything?
  3. Can this run on my infrastructure?
  4. Who else can see it?
  5. What happens to it if we stop working together?

If a provider hesitates on any of these, or answers in jargon, you already have the information that matters. A serious agency answers in two clear sentences. If you want a deeper look at how to evaluate a partner, we wrote a separate guide on how to choose an AI automation partner. A few more useful questions there.


What it costs (briefly)

The 30 minute conversation is free. The specification costs 400 EUR, credited against the build if you sign within 60 days and free with an existing DIGIVEST strategy or feasibility study. A build starts at 1,500 EUR, fixed from the specification. Monthly operation for agents connected to business systems starts at 250 EUR.


Frequently asked questions

Is it safe to give my customer data to an AI?

Live agents run in isolated instances on re:solved’s server in Timișoara. Message text is sent to the OpenAI API under a data-processing agreement. Entirely local hosting and processing are design options, not descriptions of live deployments. Local hosting alone does not guarantee compliance.

Is AI automation GDPR compliant?

There's no yes/no answer. GDPR doesn't ban AI. It requires you to know where data goes, to have a lawful basis, to have a contract with your processors, and to honour your customers' rights. An AI tool is compliant when it's configured to respect all of that; it stops being compliant when it's added to your business without a thought for any of it.

Do I need to wait for the EU AI Act before using AI in my business?

No. Its transparency duty already applies: since 2 August 2026, a system that converses with people must tell them it isn't human, and that's a design choice, not a blocker. The heavier high-risk obligations apply from December 2027, after the EU deferred the original date. GDPR already applies and is the day-to-day framework for personal data.

What does "self-hosted" actually mean for an AI agent?

Our agents run on re:solved's server in Timișoara; message text goes to the OpenAI API under a data-processing agreement. Hosting on your premises is possible on request.

If I use ChatGPT or Copilot for work, am I breaking GDPR?

Not automatically, but it depends on what you paste in and which settings are active. These tools have business tiers with data-processing agreements and options not to use content for training. For personal customer data, you need those tiers plus an internal check on what your team actually types in.


The other half of the compliance question: What the EU AI Act actually asks of a business using an AI assistant

Want an honest conversation about your data?

Book a 30 minute conversation. In 30 minutes we look at one process that costs you time. We tell you honestly whether an agent can take it over and what it would cost. No sales presentation.