Quiz EN

1. Your colleague claims: "I only use our internal company tool, so the EU AI Act doesn't apply to me." What do you respond? *

Correct, internal tools are completely exempt from the EU AI Act

Wrong, the EU AI Act applies to all AI users regardless of the tool

Correct, only users of ChatGPT and similar tools are affected

The EU AI Act doesn't take effect until 2030 and isn't relevant yet

2. An LLM generates a sentence with a specific year and source citation. Why is this NOT proof of accuracy? *

LLMs generate statistically plausible word sequences, not verified facts

LLMs fundamentally cannot process numbers correctly

LLMs have no access to databases containing dates

3. Which statement about the "Human in the Loop" principle is FALSE? *

For decisions with legal effect, a human must make the final decision

A quick skim of the AI result fulfills this requirement

This principle is a central component of the EU AI Act

It specifies and strengthens Article 22 of the GDPR

4. You need to prepare a termination letter for an employee using AI. What is legally problematic? *

Using AI for texts in the HR area is generally prohibited

If the AI makes or influences the termination decision itself

Automatic formatting of HR documents by AI

AI may be used without restrictions in the HR area

5. You need a complex project plan. Which combination of prompting strategies is most effective? *

Combine format, context, step-by-step thinking and quality criteria

Formulate a prompt as short and concise as possible

Only use shot prompting with one detailed example

Use exclusively role prompting with a clear expert role

6. Why can a low temperature parameter reduce the hallucination risk in LLMs? *

The AI then works more slowly and makes fewer mistakes

Lower temperature leads to more focused, less creative responses

The AI then only accesses verified facts

The temperature parameter only affects response speed

7. You post an AI-generated, photorealistic image of a fictional product on LinkedIn. What does the EU AI Act require? *

No labeling required since it's marketing

A simple copyright notice at the image edge is sufficient

The image must be deleted after 24 hours at most

Labeling requirement since the image is not recognizably AI-generated

8. What is the essential difference between "confabulation" and deliberate lying in AI? *

There is no essential difference between these two phenomena

Confabulation is unconscious gap-filling without intent to deceive

Confabulation only occurs in image generation

Deliberate lying occurs significantly more often in AI systems

9. Your team wants to establish an AI workflow. Which step is missing from this list: "Define AI use → Use result → Document"? *

Review and validate result by a human

Add additional AI tools for security

Fully automate the entire workflow

Acquire an additional AI tool before starting

10. Which statement about AI labeling is CORRECT? *

Only AI-generated images must be labeled

Internal company documents are generally exempt

Texts must also be labeled when used externally

Labeling is only a recommendation, not mandatory

11. Why is "Grounding" an effective strategy against hallucinations? *

The AI then automatically switches to English responses

The AI relies on provided info instead of its "memory"

Grounding completely deactivates the AI's creative functions

This strategy only works for image generation

12. A chatbot on your website advises customers on insurance products. Which EU AI Act obligation is PRIMARILY relevant here? *

All images used must be marked as AI-generated

Users must be able to recognize they are talking to an AI

The chatbot source code must be publicly accessible

Chatbots in customer service are exempt from the EU AI Act

13. When does even a perfectly formulated prompt reach its limits? *

Good prompts always lead to the desired result

For knowledge after the training cutoff and complex value judgments

Exclusively for mathematical calculations and formulas

Only when the prompt exceeds a certain character count

14. You adopt an AI-generated report unchecked and it contains a factual error that leads to losing a customer. Who is liable? *

Liability lies entirely with the manufacturer of the AI system

Nobody, as AI-generated content is not legally covered

You as the user, since you are responsible for AI content you use

The IT department that provided the AI tool

15. What best describes the paradigm shift from "AI as a tool" to "Human-Centered AI"? *

AI systems are becoming significantly faster through new technologies

AI will make all important decisions completely autonomously in the future

Humans remain the decision-makers, AI supports without replacing

AI will only be used for image generation in the future