A representative survey by TU Darmstadt of more than 2,000 people in Germany has found a striking and counterintuitive pattern: the better people understand artificial intelligence, the more they expect it to take their jobs. Around 43% of respondents with very strong AI knowledge said they believe AI could soon take over their work, a higher share than among those with less understanding. The finding challenges the comforting assumption that familiarity with AI reduces anxiety about its impact on employment.
Key facts from the survey
- 43% of respondents with very strong AI knowledge expect AI to take over their work.
- Only 15% of all respondents have had any AI training.
- The survey was carried out by TU Darmstadt with YouGov and is representative of the German population.
- Concerns about hallucinations and data protection rank higher than fears of superintelligence.
- Younger workers are more anxious about job security than older workers.
- Job security concerns rose across every occupational group compared with last year.
A surprising pattern of fear
Peter Buxmann, who led the study at TU Darmstadt, described the results as both remarkable and worrying. His team found that people who understand AI best are also the most aware of its disruptive potential. This is not a story of ignorance breeding fear. Instead, it suggests that deeper knowledge brings a clearer view of how fundamentally AI could reshape workplaces and careers.
The research was conducted in Germany with YouGov, a market research firm, and the results were measured against an earlier edition of the same survey. Because the sample is representative, the findings offer a reliable snapshot of how German workers feel about AI at a moment of rapid technological change.
Output matters more than industry
One of the most important insights from the survey is that traditional industry categories are not the best predictor of AI risk. What matters more is the nature of the work itself. Jobs that involve producing text, analysis, reports, code or slides are considered more exposed to AI-driven automation. This shifts the discussion away from manufacturing versus services and toward a more granular look at daily tasks.
For example, a marketing manager who spends hours drafting reports may feel more vulnerable than a warehouse worker performing physical tasks. Similarly, a software developer who writes code may worry more than a nurse who provides hands-on care. The survey indicates that white-collar and knowledge-based roles are now at the center of AI-related job anxiety.
Concrete worries outweigh superintelligence
The specific fears expressed by respondents are also revealing. Rather than worrying about machines becoming superintelligent or taking over humanity, people are focused on more immediate and unglamorous problems. Hallucinations, where AI generates confidently incorrect information, and data protection concerns top the list of worries.
Only 40% of respondents think that superintelligence is even likely. This suggests that most people are not losing sleep over science-fiction scenarios. They are instead concerned with the practical consequences of using AI systems that can produce misleading outputs or mishandle sensitive data. These are issues that employers and policymakers can address with better training and clearer safeguards.
Younger workers are more anxious
Another surprising result is that younger workers are more anxious about AI and job security than older workers. This inverts the usual expectation that older employees would fear technological change more. The reasons may include the fact that younger people are more likely to work in roles that require digital skills and produce analyzable output. They may also have grown up in an era of rapid tech disruption and therefore feel less certain about long-term career stability.
The survey found that concern about job security rose across every occupational group since last year. This indicates that AI-related anxiety is not limited to a specific sector or age bracket. It is becoming a pervasive feature of the German labor market, and likely of other advanced economies as well.
The AI training gap
Perhaps the most troubling number in the entire survey is that only 15% of respondents have had any AI training at all. This is striking because Europe has had a rule about AI literacy for eighteen months. Article 4 of the European Union's AI Act has applied since 2 February 2025. It obliges providers and deployers of AI systems to take measures supporting AI literacy among staff who operate those systems.
Article 4 is not a rigid mandate with a fixed curriculum. It sets no particular level of knowledge and binds only organizations that actually use AI systems. Therefore, the 15% figure is not a direct measure of legal compliance. It is more accurately a readiness figure. It shows how far most German workers are from having the basic understanding that the AI Act envisions.
Eighteen months after the literacy duty began to apply, most workers across Germany have received no training. This gap is deeply concerning for Buxmann, who argues that training and honest answers are needed rather than reassurance. Losing a job costs more than income, he points out, because employment is tied to identity, social connection and mental well-being.
What is the EU AI Act?
The EU AI Act is a comprehensive regulatory framework for artificial intelligence, designed to ensure that AI systems placed on the European market are safe and respect fundamental rights. It establishes a risk-based approach, with different obligations depending on the level of risk posed by an AI system. Article 4 focuses specifically on AI literacy, requiring providers and deployers to ensure that their staff have the necessary skills to use AI systems responsibly.
This includes understanding the capabilities and limitations of AI, as well as knowing how to detect and address issues like bias and hallucinations. The provision is intended to create a baseline level of awareness across the workforce, so that people who interact with AI systems are not blindly relying on them. Yet the survey shows that this goal is far from being achieved.
The role of employers
The findings place a significant responsibility on employers. Companies that deploy AI systems must do more than simply buy software and hand it to employees. They need to invest in continuous learning and support. Without training, workers cannot understand what AI can do, what it cannot do, or how to challenge its outputs. This lack of understanding amplifies fear and makes it harder for employees to adapt to changing roles.
Training should not be a one-time event. It needs to be ongoing, updated as AI technology evolves. It should also be practical and role-specific, focusing on real tasks and real risks. For example, someone who uses a chatbot to answer customer queries needs to know how to verify accuracy and protect data. A developer who uses code-generation tools needs to understand how to evaluate the security and reliability of generated code.
Honest communication is needed
Buxmann’s call for honest answers rather than reassurance suggests that organizations often downplay the risks of AI in an attempt to calm employees. But vague reassurance rarely works. Employees know when their jobs may be affected, and they can see the rapid advances in AI capabilities. What they need is transparent communication about how AI will be used, what tasks may change, and what skills will be valuable in the future.
Employers should also involve employees in decisions about AI deployment. When workers have a voice in how technology is introduced, they feel more control and less likely to see AI as an existential threat. This participatory approach can reduce anxiety and improve the quality of AI adoption.
The survey also suggests that industry and policy leaders should pay attention to the fact that younger workers are more worried. These are the people who will form the core of the future workforce. If they feel threatened by AI, it could affect their career choices, their willingness to invest in skills, and their overall economic outlook. Proactive measures, including education and social safety nets, are essential.
In the end, the TU Darmstadt survey paints a picture of a workforce that is both informed and anxious. Those who understand AI best are not complacent; they are worried. The only 15% AI training rate is a wake-up call for Europe’s AI literacy ambitions. The numbers are a reminder that laws alone do not create understanding. Real progress depends on employers, governments and educational institutions taking the task of AI literacy seriously, not as a checkbox but as a strategic priority.