Privacy & security · 6 min read

Using AI within the GDPR.

AI tools save a lot of time, but what happens to the data your employees paste into them? In this article you read which rules apply and which practical measures help you use AI safely.

Bastiaan Meijer · CareScript

The GDPR also applies to AI

As soon as personal data goes into an AI tool, the GDPR applies. Think of a customer email being summarised, a CV being assessed or meeting notes containing names. The familiar principles still apply:

  • Lawful basis: you need a valid reason to process the data.
  • Purpose limitation: only use data for the purpose for which it was collected.
  • Data minimisation: do not share more than needed. Often it works without names too.
  • Transparency: tell the people concerned how you use their data.

Consumer version or business version?

The biggest risk often lies in using free consumer versions. What you enter may be used to improve models, and you have no data processing agreement. So choose business versions or API access, where:

  • you sign a data processing agreement;
  • your data is not used for training;
  • it is clear where the data is processed and how long it is kept;
  • you have control over users and access.

Data outside the EU

Many AI services process data (also) outside the European Union. That is allowed, but only with appropriate safeguards, such as an adequacy decision or standard contractual clauses. Check this before rolling out a tool across your organisation, and choose processing within the EU where possible.

When is a DPIA needed?

Do you use AI in a way that could pose a high risk to people, for example when assessing job applicants, processing health data or large-scale analysis of behaviour? Then a data protection impact assessment (DPIA) is often mandatory. It maps out the risks in advance.

And the EU AI Act?

Alongside the GDPR there is the EU AI Act. It works with risk levels: the higher the risk of an application, the stricter the requirements. A few applications are prohibited, strict requirements apply to high-risk applications and for most everyday applications the rules are mainly about transparency.

One obligation applies to almost every organisation that uses AI: making sure employees who work with AI are sufficiently AI literate. They need to understand what the tools can do, where the risks lie and how to check results.

Five practical measures

  1. Set out in a short AI policy which tools are allowed and which data may and may not go into them.
  2. Only use business accounts with a data processing agreement.
  3. Remove names and identifiable data where possible.
  4. Always let a person take and check important decisions.
  5. Train your employees, so they use AI safely and effectively.

This article is a practical introduction and not legal advice. If you are unsure about your situation, consult your data protection officer or a lawyer.

Further on this topic

AI training for teams

Read how CareScript can help you with this, or book a no-obligation call straight away.