Artificial intelligence has long since become part of everyday life. But there’s a big difference between a prompt and true AI expertise. What people should know about AI—and why the AI Act is now even addressing the issue from a legal perspective.
As of September 3, 2026 · This post is for general informational purposes only and does not constitute legal advice.
Using artificial intelligence has become surprisingly easy.
Type in a question.
Press Enter.
Get an answer.
You can summarize texts, generate images, prepare presentations, draft emails, or develop ideas. You don't need to have studied computer science or understand how a neural network works to do this.
And that is precisely where one of the great strengths of modern AI lies.
At the same time, however, this gives rise to an interesting misunderstanding:
- Just because we can use AI doesn't mean we understand it.
After all, just because someone knows how to use Google doesn't mean they automatically understand how a search engine works. Just because someone drives a car doesn't make them an automotive engineer.
When it comes to artificial intelligence, however, there is one crucial factor to consider:
- We use systems whose results sometimes seem so convincing that we can easily forget to question them.
And suddenly, a technical issue becomes an educational one.
01 — WHAT DOES AI LITERACY ACTUALLY MEAN?
In connection with the European AI Act, we are now encountering a term more and more frequently: AI literacy, usually referred to as “KI-Kompetenz” in German. The AI Act defines AI literacy as the skills, knowledge, and understanding that, among other things, enable the informed use of AI systems and are intended to raise awareness of opportunities, risks, and potential harm. AI Act Service Desk
That's remarkable. Because it's not just about which tool someone can use.
AI literacy means, for example, being able to assess:
What can an AI system do?
What can't it do?
Where might errors come from?
How reliable is a result?
What data do I provide to the system?
What impact might its use have on other people?
And perhaps one of the most important questions of all:
- When should I not just take an AI at its word?
For me, this is exactly where true AI expertise begins. Not with the perfect prompt, but with an understanding of what happens before and after that prompt.
02 — WHY THIS TOPIC HAS SUDDENLY APPEARED IN THE AI ACT AS WELL
In the European AI Act, AI literacy is not merely a lofty educational goal.
It has been given its own section: Section 4.
The regulation has been in effect since February 2, 2025, and was amended in 2026 by the Digital Omnibus. According to the current version, providers and operators of AI systems must take measures to support the development of AI competence among employees and other individuals who operate or use AI systems on their behalf. In doing so, they must take into account, among other things, prior knowledge, experience, training, and the specific context of use. EUR-Lex
However, that does not mean:
- Every employee must now obtain a standardized AI certification.
Nor does the AI Act prescribe a specific course or a specific number of training hours. Context, in particular, plays a role. An employee who occasionally uses AI to draft internal documents needs different knowledge than someone who uses an AI system to evaluate job applicants or prepare technical decisions. And that is exactly the point.
AI literacy cannot mean the same thing to everyone.
03 — A PROMPT DOESN'T EQUAL COMPETENCE
In public discourse, however, AI literacy is surprisingly often reduced to the use of generative AI.
Then it's about:
“The 10 Best ChatGPT Prompts.”
“Use These Five Tricks to Use AI Like a Pro.”
“How to Automate Your Workday.”
This can be useful. But it’s just a small part of the picture. Let’s imagine someone who has mastered all prompt techniques perfectly. That person can create impressive texts, generate images, and conduct research using AI. But they don’t realize that a language model can produce convincing-sounding misinformation. They don’t verify sources. They copy confidential company information into an unverified system. And they incorporate unfiltered AI results into important decisions.
Would we really describe this person as AI-savvy?
Probably not.
Using a tool is simply not the same as understanding a tool.
04 — AI COMPETENCE HAS SEVERAL LEVELS
To use AI responsibly, we therefore need several types of knowledge.
Understand
People don’t need to know every mathematical detail of an AI model. But a basic understanding helps. A language model doesn’t automatically “know” what is true or false. It generates output based on learned patterns and probabilities. That’s a crucial difference.
Apply
Of course, practical application is also part of it. How do I formulate meaningful tasks? What information does the system need? Which AI tool is actually suitable for my purpose?
Check
Perhaps this very skill is becoming increasingly important. Is the answer correct? Can I verify the source? Is there a lack of context? Is the result possibly skewed? An impressively worded answer is not proof of its accuracy.
Identifying Risks
Data protection, copyright, discrimination, manipulation, inaccurate information, or undue reliance on automated results: AI brings more than just opportunities. Competence, therefore, also means being able to identify and assess risks.
Take Responsibility
And finally, there remains a question that already played a central role in the first article of this series:
Who makes the final decision?
AI can support people. But the greater the impact of a decision, the more important it is for people to be able to understand, review, and, if necessary, challenge the outcome.
05 — AND WHAT DOES THIS HAVE TO DO WITH EDUCATION?
Quite a lot. After all, AI literacy doesn’t begin the moment someone signs an employment contract. Children and teenagers are growing up in a world where AI will be a given. They will use these systems to learn, conduct research, write, program, and engage in creative work.
That is why, in my view, it would not be enough to simply show them:
- "Here's an AI tool. Here's how to use it."
The much more exciting task is:
- How can we help young people understand what they're using?
And to do that, they need skills that, at first glance, don't sound anything like AI:
logical thinking,
problem-solving,
critical thinking,
evaluating sources,
recognizing connections,
finding errors,
developing one’s own solutions,
Interestingly, these are many of the same skills that play a role in learning to program. And that is precisely why I do not view programming and AI literacy as two competing educational goals. On the contrary. Anyone who experiences, while programming, that a computer does exactly what it was told to do—and that an error must be identified, understood, and corrected—develops a different perspective on technology. Not one of awe. Not one of fear, but one of curiosity and critical thinking.
06 — DOES EVERYONE NEED TO UNDERSTAND HOW A NEURAL NETWORK WORKS NOW?
No. That would be just as unrealistic as requiring everyone to be able to build a processor before using a computer. AI literacy should therefore not mean turning everyone into an AI developer. It’s about contextual knowledge. A student needs different knowledge than a human resources manager. A teacher needs different knowledge than a software developer. A CEO needs different knowledge than someone who occasionally uses AI for personal writing. The AI Act also takes this contextual approach into account in Article 4. EUR-Lex
Therefore, the goal cannot be:
- Everyone needs to know everything about AI.
Rather:
- People should understand enough to be able to meaningfully evaluate AI in its specific context.
07 — WHAT MIGHT AI LITERACY LOOK LIKE IN PRACTICE?
There is no single mandatory training program for this.
The European Commission is now compiling practical examples from various organizations. Its “Living Repository” contains more than 40 initiatives—ranging from e-learning courses and in-person training to boot camps and collaborations between academia and industry. The Commission expressly points out that simply copying these examples does not automatically mean that Article 4 has been fulfilled. Europe’s Digital Future
This also shows that:
- AI expertise isn't just a box to check on a checklist.
A useful approach could, for example, begin with a very simple question:
- What kind of AI do we actually use—and for what purposes?
After that, things will become more specific.
What opportunities does it offer?
What errors can occur?
What information can be entered?
When do results need to be reviewed?
Who might be affected by these results?
And who is ultimately responsible?
Suddenly, an abstract concept becomes something very practical.
08 — MAYBE THAT'S THE REAL CHALLENGE
We're talking an incredible amount right now about what AI will be capable of. Which jobs it will transform. Which tasks it will automate. Which models will become more powerful. Which companies are leading the way.
But perhaps we ask the more important question far too rarely:
- How advanced do we humans need to become in order to use increasingly powerful systems effectively?
Because, to me, technological expertise doesn't mean chasing after every new development. Nor does it mean having to try out every new tool right away. And it certainly doesn't mean either unconditionally praising AI or rejecting it outright.
It means:
understand,
try out,
question,
classify,
decide.
Perhaps that is precisely one of the most important educational tasks of our digital age.
09 — NOT JUST AI. MORE UNDERSTANDING.
The coming years will likely bring us many new models, applications, and possibilities. Some tools will disappear. Others will be added. User interfaces will change. Features that seem spectacular today will eventually be taken for granted. That’s why I believe that focusing solely on tool proficiency is too short-sighted. Anyone who has only learned where each button is located will have to start from scratch with the next tool. Those who, on the other hand, have learned how to evaluate technology will carry that knowledge with them.
And for me, this brings us full circle back to education and programming:
- The point isn't to teach children today exactly the software they'll be using in ten years.
We probably don't even know this software yet. The point is to equip them with the skills they need to navigate a digital world that we don't yet fully understand. And ultimately, the same applies to adults.
AI literacy, therefore, does not begin with the question:
- What prompt do I need to enter?
But with a much more important one:
- Do I understand the system well enough to decide what to do with his answer?
Perhaps that is exactly the difference between using AI and understanding AI.
Sources
The current text of the AI Act, particularly Article 4 and the definition of AI competence, as well as information from the EU AI Office and the European Commission on AI literacy and the Living Repository. EUR-Lex
AI Act – Current Text of the Regulation on EUR-Lex
European Commission: AI Literacy, Skills, and Talent
European Commission: Repository of AI Literacy Practices