What Does the Obligation to AI Competence Mean?

What is Required

Employees using AI systems should be able to assess their functionality, limitations, and risks – in relation to their role and the deployed system.

For a text suggestion function, less understanding is required than for a system that prepares decisions.

What This Means Practically

Not: a certification or a multi-day course for everyone.

But: People should know,

• what the system does and what it does not
• that it can make mistakes and what that looks like
• which data they are allowed to input and which they are not
• when they need to verify the results
• whom to contact in case of doubts

A Practical Approach

1. List which AI systems are used in-house – including those that have not been registered by anyone.
2. A brief document for each system: purpose, limitations, allowed data.
3. A training session, documented.
4. Repeat for new systems.

Why This Makes Sense Even Without an Obligation

The most common damages arise not from poor models, but from false expectations: Someone relies on a fabricated statement or inputs confidential data into a private account. Both prevent clarification more effectively than any technical measure.

Documentation

Document who was trained on what and when. Without documentation, compliance cannot be proven.

This text does not constitute legal advice.