Digee / Wiki / Law and security
Overview
AI and GDPR: what a business must watch out for.
GDPR does not ban using AI with personal data. It does require you to know what data the AI processes, why, where and who has access to it. With a well-designed AI agent, these questions are answered right at the start.
In short
- Using AI with personal data is lawful if you have a legal basis and a contract with the provider.
- Give AI only the data it genuinely needs for the task.
- Do not leave decisions with a serious impact on people purely to automation.
On this page
Five questions to ask
- What personal data does the agent process? Names and email addresses? National ID numbers? Health data?
- On what legal basis? Most often performance of a contract, a legal obligation or legitimate interest (Art. 6 GDPR).
- Who is the processor? The AI service provider and the agent supplier both process data on your behalf – you need a data processing agreement with them (Art. 28).
- Where is the data processed? Inside or outside the EU? A transfer outside the EU needs a legal framework.
- How long is the data kept and who has access to it?
Contracts with AI providers
Major providers of language models offer terms for business and API use in which they commit not to use customer data to train their models, along with data processing addenda (DPAs). This often does not apply to free versions of public chat tools.
A practical rule: do not paste clients' personal data into free public chat tools. For business use, choose services with a contract and a clearly defined place of processing.
Transferring data outside the EU
Many AI services run on servers in the USA. A transfer is possible, for example, under the EU–US Data Privacy Framework if the provider is certified, or under standard contractual clauses. Many providers now also offer processing in EU data centres. When designing an agent, it is worth making this decision deliberately.
Automated decision-making
Article 22 GDPR gives people the right not to be subject to a decision based solely on automated processing that has legal or similarly significant effects on them (e.g. rejecting an application, selecting a job candidate). That is why, for such steps, the agent prepares the groundwork and a person decides.
When a DPIA is needed
A data protection impact assessment (Art. 35) is required where processing is likely to result in a high risk to people's rights. With AI, this typically applies to large-scale processing of sensitive data (health), systematic evaluation of people, or the use of new technologies on a large scale. Simple sorting of business mail usually does not require one, but it is good practice to document that conclusion.
A practical approach to deploying an agent
- Describe what data the agent sees and why.
- Limit access to the folders and systems it actually needs (data minimisation).
- Sign data processing agreements with suppliers.
- Choose where processing happens (EU / outside the EU) and document the legal framework.
- Update your records of processing activities and the information given to data subjects.
- Decide what a person must approve.
The supervisory authority in the Czech Republic is the Office for Personal Data Protection (ÚOOÚ). We cover the link to the EU regulation on AI in our article on the AI Act.
This article is a general overview, not legal advice. For sensitive processing, we recommend consulting your data protection officer or a lawyer.
Frequently asked questions
Can I paste a client's email into ChatGPT?
Better not into the free public version if it contains personal data. For work with client data, use a business version or a service with a data processing agreement and a clearly defined place of processing.
Do AI providers use my data for training?
With the business and API versions of the major providers, usually not – this is part of the contract terms. Free consumer versions may be different. Always check the contract.
Do I need a data protection officer because of an AI agent?
The obligation to appoint a data protection officer does not arise from AI itself but from the nature of your activities (e.g. public authorities, large-scale processing of sensitive data).
Want to know where AI would save you time?
We go through a normal week with you and find the first task worth handing to an agent. No obligation.
Related articles
AI Act: what the EU regulation on AI means for businesses
The EU Artificial Intelligence Act in plain English: risk categories, application dates including the 2026 delay, obligations for businesses and fines.
→Explainer · 3 minHuman-in-the-loop: a person decides
Human-in-the-loop means AI prepares the work and a person approves it. How to set an agent’s limits so it saves time without making costly mistakes.
→Explainer · 3 minRAG: how to teach AI your company documents
RAG is a way for AI to answer from your documents instead of its general memory. How it works, what it is useful for in a business and what to watch out for.
→