Skip to content
Step 1 / –
flag Introduction

Checking the result

AI can make mistakes without giving you any warning. So check what you send or use as the basis for a decision. In this module you will learn what to look out for.

target What you will learn in this module
  • check_circle What you always check before you use AI text
  • check_circle How bias creeps into AI answers
  • check_circle Why genuine human review is needed
draw You sign off

You sign off, not the AI

AI sometimes invents figures, quotes, legal articles or sources that do not exist. This is called Hallucinating An AI tool stating something with great confidence that is not true: a made-up figure, a quote that appears nowhere or a source that does not exist. It happens with all AI tools, even the best ones. . The answer sounds just as convincing as when it is right.

So always check:

  • figures and calculations: recalculate them or compare with the source;
  • names, dates and quotes: are they really in the original document?
  • legal or tax statements: have them checked by someone who knows;
  • tone and content: would you write this yourself, like this, to this customer or colleague?
balance Bias

Watch out for bias

AI can learn from data that contains bias, such as texts from the internet. That bias can end up in the answers without anyone noticing.

This is what it can look like:

  • In job adverts: AI mainly suggests men for technical jobs and women for care jobs.
  • With candidates: a CV with a certain name or home town gets a lower score.
  • In texts: examples and images keep showing the same kind of people.
balance What you do

Read critically, especially when the answer is about people. Ask yourself: would I decide the same if the AI had not suggested this?

groups Deciding about people

Make sure there is genuine human review

The practical rule for employees: do not take the initiative to let AI decide about people on its own. Use only an application that has been assessed and approved beforehand, with genuine human review. Think of:

  • selecting or rejecting candidates;
  • assessing employees’ performance;
  • refusing a customer or setting a price based on a profile.

AI can help within an approved application, for example with CV summaries. The reviewer must be able to check the relevant source information, have sufficient knowledge and authority, and be able to reject the AI’s suggestion. Just clicking “approve” is not genuine review.

The GDPR restricts fully automated decisions with legal effects or similarly significant effects. There are exceptions, with conditions and safeguards. You do not apply those yourself: have such an application assessed beforehand by the person responsible within your organisation. Even with human review, using AI in recruitment is not automatically permitted.

gavel No emotion recognition

Using AI to infer emotions at work or in education from biometric data, such as facial or voice characteristics, is prohibited in the EU, except for medical or safety purposes. Analysing a text for tone or sentiment does not fall under that specific prohibition. That does not mean every use of it is permitted: do not use such tools on colleagues, candidates or customers on your own initiative.

quiz What would you do?

A summary that is not right

description
You have AI summarise a 20-page contract. The summary mentions a notice period of three months, but you cannot find that clause anywhere in the contract.

What do you do?

quiz Practice · question 1 of 2
info Just practice, this does not count toward your progress
AI writes a quote with a discount you never agreed. What do you do before you send it?
Correct: AI is a tool. Whatever you send or decide on the basis of AI, you check yourself.
quiz Practice · question 2 of 2
What may AI do in a recruitment process?
Correct: first, this application has to be assessed and approved. An authorised reviewer checks the relevant source information, weighs up candidates themselves and can reject the AI's suggestion.
summarize Summary

What you take away from module 4

  • bolt AI sometimes invents facts, quotes and sources, and can sound confident while doing so.
  • bolt Check figures, names, quotes and legal statements before you use anything.
  • bolt AI can pick up bias from the data it learned from. Read critically.
  • bolt Use AI in decisions about people only within an approved application with genuine human review.
  • bolt Biometric emotion recognition at work is prohibited, with medical and safety exceptions. Do not use such tools on your own initiative.
workspace_premium Module complete

Module 4 complete 🎉

You now know how to check the result of AI. In module 5 you will learn when to say that AI helped write something, and how to handle AI that records meetings.

lock_open 4 of 8 modules

On your way to the “Using AI safely” practice quiz

Complete all 8 modules and try the free practice quiz. The official exam with a certificate is in the GDPRWise app.

check_circle Modules 1-4 completeradio_button_unchecked Modules 5-7radio_button_unchecked Final exam ≥ 70%
workspace_premium