The EU AI Act has been in force since February 2025 – the world’s first comprehensive AI regulation. Since then, questions from universities and educational institutions have been piling up: Does this affect us? Do we need to have our AI tools audited? What happens if we do nothing? The short answer: it depends very much on what you do with AI. The longer answer is below.
The Timeline: When Does What Apply?
The AI Act was introduced in phases. Since February 2025, systems with unacceptable risk have been banned – including AI used for social scoring or manipulation of behaviour. Since August 2025, requirements have applied to general-purpose AI models (GPAI) such as large language models like ChatGPT or Gemini. From August 2026 (possibly extended to December 2027 via the Digital Omnibus), rules for high-risk AI systems will be added – and this is where it becomes relevant for educational institutions.
Four Risk Levels – and Who Is Really Affected
The AI Act distinguishes four risk levels: unacceptable risk, high risk, limited risk, and minimal risk. Most AI applications in day-to-day university operations – automated scheduling suggestions, smart archive search, chatbots for standard queries – fall into the minimal or limited risk category. They are barely touched by regulation.
High-risk systems in the education sector (Annex III, Point 3) are those that are used in admissions decisions, exam assessments, or learning performance evaluations and that make those decisions substantively – not merely support them. Anyone who purchases or develops an AI system that makes such decisions must classify it as a high-risk system and meet the corresponding requirements: transparency, documentation, human oversight, and technical robustness.
Am I affected as a student? In the vast majority of cases, no: students are considered end users – not operators or providers. The AI Act is primarily directed at the institution that deploys AI systems.
Am I affected as a lecturer? Again, mostly no: anyone using AI as a tool for lesson preparation or delivery is not classified as the operator of a high-risk system – as long as the final decision remains with a human.
Why TraiNex Carefully Considers Risk Levels
At TraiNex, we make a clear distinction with every new AI feature: does the function support a person in making a decision – or does it make the decision itself? AI-assisted timetable suggestions, automatic attendance list evaluations, or intelligent document archive search functions support administrative staff without replacing decisions. That is the space in which we operate – deliberately and thoughtfully.
Outlook: AI Agents and the AI Act
What about the new TraiNex AI agents that operate autonomously as digital staff members within TraiNex and make decisions? Yes, these AI agents are very powerful – and depending on the use case, not automatically compliant with the AI Act. Each agent is therefore specially trained, and every action is traceable (Explainable AI). The agents thus remain within the boundaries of what is permitted under the AI Act.
Conclusion
The EU AI Act is not cause for panic, but it is a reason to reflect. Those who use AI only as a tool for support – and leave the final decision to humans – are on the right side of the regulation. Anyone planning larger AI projects should check early on which risk category they fall into.
PS: Obligation to Promote AI Literacy Has Applied Since February 2025
Article 4 of the AI Act requires all universities that use AI to ensure that their staff have sufficient AI competence – and to document this. This applies regardless of the risk level of the system in use. TraiNex supports you with this: our e-tutorial module “AI at the University in 10 Chapters” lets you store a certified AI training programme for teaching and administrative staff directly in TraiNex, have it completed, and automatically document the evidence. Get in touch with us.