KI-Governance

EU AI Act from 2 August 2026: What Universities Need to Know

Claudia

From 2 August 2026, further key provisions of the EU AI Act apply. Does this affect us as a university? Do we now need to have our AI tools reviewed? What happens if we do nothing? The short answer: it very much depends on what you do with AI. The longer answer is here.

Lecturer and student looking at digital AI icons with EU map in the background – EU AI Act and universities
Illustration: AI-generated image (Dr. Maex / TraiNex) – EU AI Act and universities in the digital age

What Applies from 2 August 2026?

Anyone who interacts with an AI system must in principle be able to recognise that they are dealing with AI. A chatbot for students or a digital assistant should therefore identify itself clearly as an AI system at the latest at the beginning of the conversation – unless this is already obvious.

For universities, this means above all:

  • Clearly label AI chatbots,
  • Do not present AI-generated media as authentic recordings,
  • Define responsibilities for AI systems in use,
  • Document AI deployment in a traceable manner.

AI Literacy Remains Mandatory

Since February 2025, institutions have been required to offer measures to promote the AI competence of their staff. From August 2026, compliance with this obligation can be more strictly monitored.

A specific certificate is not required. However, the following are advisable:

  • Internal rules for ChatGPT and other AI tools,
  • Training on data protection, hallucinations and confidential data,
  • Special briefings for sensitive areas such as the examinations office, admissions and HR,
  • Documentation of the measures offered.
  • TraiNex supports you with the e-tutorial “AI at the University in 10 Chapters”. You can store this directly in TraiNex for teaching staff and administrative employees, have them complete it and automatically document the evidence. Contact us.

Are Students Affected?

When using generally available AI tools privately, students are normally neither providers nor operators within the meaning of the AI Act. They are, however, protected as data subjects. If a university uses AI for admissions, examination assessment or examination monitoring, obligations arise for the university.

What Applies to Teaching Staff?

Teaching staff are likewise neither providers nor operators within the meaning of the AI Act. However, they must comply with the university’s requirements, data protection law, copyright law and examination regulations. In order to assess the use of AI in these areas, participation in further training is necessary.

How TraiNex Classifies AI Functions

At TraiNex, we assess every AI function based on its specific purpose and impact.

AI-supported search functions, text suggestions, summaries and general chatbots assist users without themselves making decisions about admissions, grades or sanctions. These risk classes are largely permitted without issue.

The picture may look different when an AI result is used directly for consequential decisions. An automatic evaluation of attendance data, for example, is to be assessed differently if examination eligibility is derived from it without further review.

That is why we specifically examine:

  • Purpose and data used,
  • Impact on students and staff,
  • Degree of automation,
  • Human control options,
  • Logging and traceability,
  • Data protection and IT security.

What Applies to AI Agents?

AI agents do not form a separate risk category in the AI Act. Here too, what matters is which tasks they take on.

An agent that prepares documents, transfers information or proposes changes for approval is to be assessed differently from an agent that independently rejects admissions, changes grades or sanctions individuals. The latter would be too autonomous and too risky under the AI Act. To achieve an acceptable and permitted risk class, at minimum human approval would be mandatory.

At TraiNex AI agents, we therefore rely on:

  • clearly defined areas of responsibility,
  • graduated access rights,
  • complete logging,
  • human approvals for consequential actions,
  • the ability to stop or reverse actions,
  • restriction of AI autonomy.
    See waas.campus-management-system.de

What Universities Should Do Now

Universities and educational institutions should now:

  1. Record the AI systems they use,
  2. Document their specific purposes of use,
  3. Correctly label AI chatbots and generated content,
  4. Offer staff appropriate training and guidelines,
  5. Particularly scrutinise applications that influence admissions, examinations or HR decisions.

Conclusion

From 2 August 2026, transparency, labelling and documented accountability will become more important for universities above all. Next milestone: the comprehensive high-risk requirements for the education sector apply from 2 December 2027.

Governing AI Prompts at the Institutional Level: New Paper by TraiNex at EDULearn 2026

Claudia

How can a university ensure that AI tools are used reliably, pedagogically meaningfully, and consistently — even by staff without prompt expertise? This question was at the heart of a paper presented by researchers from Trainings-Online GmbH and TH OWL at the 18th international EDULearn Conference in July 2026 in Palma.

Conference poster: Prompt Governance in Higher Education (EDULearn 2026)
Conference poster from EDULearn 2026 — click to enlarge

The paper “Prompt Governance in Higher Education: Curated Prompt Libraries for Reliable AI Assistance” introduces a curated prompt library as a governance approach — directly integrated into a campus management system. Rather than leaving AI usage to chance, universities can centrally manage prompts, control target audiences, and monitor usage.

The system distinguishes five prompt types: from freely editable prompts to database-based queries to agent-driven workflows that autonomously conduct literature searches in external databases. Embedded in a campus chat assistant, the system identifies the user’s intent and automatically selects suitable prompts — including contextual enrichment from module handbooks and course schedules.

The approach demonstrates: reliable AI in higher education does not require complex infrastructure — it requires smart integration into existing systems like TraiNex.

→ Read the full paper on ResearchGate

Intellectual Sovereignty: What Germany’s Science Council Demands – and What We’re Already Building

Claudia

On July 6, 2026, Germany’s Wissenschaftsrat (Science Council) published its recommendations “Intellectual Sovereignty: Recommendations for Higher Education in Times of Generative AI” — nine concrete calls to action for universities, faculty, students, federal states, and the national government. Reading through them, one thought kept recurring: this is exactly what we’re building with TraiNex and SMARTA.

Intellectual Sovereignty and Generative AI – Wissenschaftsrat Recommendations 2026
Intellectual Sovereignty vs. Generative AI — click to enlarge

What the Science Council is calling for

“Critical thinking cannot be delegated to an AI,” says Council chair Wolfgang Wick. This isn’t technophobia — it’s precision. It defines what AI should do, and what it must not. The Science Council calls on universities to build AI competencies while also anchoring AI-free zones in curricula. Assessment formats should be rethought, universities preserved as social learning environments. And: sovereign AI infrastructures should be developed that work across institutions and are financed on a long-term basis.

Not as much AI as possible — but the right AI

Not as much AI as possible, but AI deployed deliberately — with the right context, for faculty, students, and university administration. For us, this means advancing context engineering: the intelligent connection of AI to each individual user’s current situation. Automatic consideration of academic records, curricula, exam regulations, timetables, and today’s lecture content. To achieve this, we connect generative AI with our algorithmic intelligence, directed by agents.

Context Engineering as the answer

An AI that automatically considers a student’s academic standing, their curriculum, their exam regulations, and today’s lecture content is not an uncontrolled black box — it’s a precise tool with a defined context. That is the core of SMARTA: not generative AI that simply responds, but AI that draws the right context from the campus management system. Study progress, curricula, exam regulations, timetables, today’s lecture notes — all of this defines the frame within which the AI operates.

Conclusion

The Science Council has set an important framework. For universities that want to deploy AI deliberately — with control, context, and competence — TraiNex shows what that looks like in practice.

→ Full press release by the Wissenschaftsrat

Machine AI vs. Human Algorithms

admin

In 1956, at the first AI conference, there was a consensus that machines could simulate intelligence. The debate was whether this would be best achieved through machine learning algorithms or human-written code. Today, many believe that AI will revolutionize university software. However, some argue that human algorithms, which have been effective and legally compliant for years, are still essential.

We believe that human algorithms are crucial for legally significant decisions, such as calculating final grades on a diploma supplement or determining a faculty’s budget. Algorithmic Intelligence, which is code written by humans, remains vital for universities. Human algorithms follow specific rules, are precise, and lack creativity. They don’t alter themselves. When a decision is made by an algorithm, it’s possible to trace back the reasoning, ensuring legal compliance. Plus, you don’t need tons of examples to train a human algorithm.

Consider a new university using generative AI to assign final grades. They would need many examples to train the model. Then, they might end up giving different grades to similar students on different days without understanding why. That would be disastrous.

Therefore, SMARTA focuses on using AI not for decision-making tasks but for areas where creativity is key, especially in generating text. We use various language models like GPT. Instead of the standard interface, we strictly utilize the API. This approach allows us to seamlessly integrate AI into existing systems like a Campus Management System and control both input and output.