If you study in the EU, put 2 December 2027 in your calendar now. That is the day the EU AI Act's high-risk rules start applying to the AI your university uses for admissions, grading, placement and exam proctoring. The rules were supposed to start in August 2026, but the EU moved them: the Digital Omnibus on AI, adopted in July 2026, pushed the date back by sixteen months. The delay gives universities time to prepare; it changes nothing about what they will owe you. This guide explains exactly what changes, which systems are affected, and what you will notice.
The short answer
- The date that matters is 2 December 2027. From then, the Chapter III high-risk rules apply to education AI in Annex III, point 3.
- The delay is already law. Regulation (EU) 2026/1744, the Digital Omnibus on AI, moved the Annex III date from 2 August 2026 to 2 December 2027 (AI in regulated products: 2 August 2028).
- Your university, not you, carries the duties. Risk management, human oversight, logging and transparency fall on the provider that builds the system and the deployer — usually your university.
- You gain a usable right. Article 86 gives you the right to a clear explanation of the role an AI system played in a decision that significantly affects you.
- Your daily study life barely changes: no ChatGPT ban, no fines on students.
The timeline: what changed in 2026
The AI Act, Regulation (EU) 2024/1689, entered into force on 1 August 2024 and applies in stages: bans and AI literacy since 2 February 2025, general-purpose AI rules since 2 August 2025, and most of the rest, including Article 50 transparency, since 2 August 2026.
The high-risk chapter was meant to follow on 2 August 2026. It did not: the Commission proposed the Digital Omnibus on 19 November 2025, the Parliament and Council agreed on 7 May 2026, and the final text was adopted on 8 July 2026 as Regulation (EU) 2026/1744, in force since 27 July 2026. Recital 40 gives the reason: the standards, guidance and national authorities needed were not ready.
| Date | What applies |
|---|---|
| 2 February 2025 | Bans (Art. 5), AI literacy (Art. 4) |
| 2 August 2025 | General-purpose AI model duties (Chapter V) |
| 2 August 2026 | Most other rules, including transparency (Art. 50) |
| 2 December 2026 | Two new bans: non-consensual intimate imagery, child sexual abuse material (Art. 5(1)(ba)–(bb)) |
| 2 December 2027 | High-risk rules (Chapter III, Sections 1–3), Annex III incl. education |
Two points matter. First, the delay covers only the high-risk rules; the bans, literacy, GPAI and transparency duties kept their dates. Second, 2 December 2026 adds two new prohibitions — AI systems that generate realistic intimate images of an identifiable person without consent, and systems that generate child sexual abuse material. Generating non-consensual intimate images of a classmate is therefore a prohibited AI practice from that date.
Which university AI counts as high-risk?
Not every AI tool on campus is high-risk. The Act names four education uses in Annex III, point 3, all “at all levels”: AI deciding access, admission or assignment; AI evaluating learning outcomes, including when results steer your learning; AI assessing the level of education you can access; and AI monitoring prohibited behaviour during tests, including webcam proctoring.
If a system does one of those four things, it is high-risk from 2 December 2027. There is one exception: systems that only perform a narrow procedural task, improve on work a human already completed, or prepare an assessment without replacing human judgement. A tool that formats a grade sheet is not high-risk; a tool that decides the grade is. But a system used to profile people in an Annex III use case is automatically high-risk — the exception does not save it.
What is not high-risk? Your own use of a chatbot to brainstorm, summarise or draft: the Act regulates the model's provider and the organisation deploying it, not the student typing a prompt.
AI detectors are not mentioned by name in the Act, and whether one used in a misconduct case counts as high-risk is unsettled — treat confident claims either way with suspicion see article #19: How Universities Detect AI Writing in 2026 — and How to Stay Honest.
One ban already applies: inferring emotions in education has been prohibited since 2 February 2025 under Article 5(1)(f), except for medical or safety reasons. A university still running such a tool is already breaking the law. The exam rules are covered in our guide to your exam rights see article #47: Your Exam Rights Under the EU AI Act: Proctoring, Emotion Recognition and the Right to an Explanation.
What your university has to do before 2 December 2027
The Act splits the work between two roles: the provider, usually a software company, builds the system; the deployer, usually your university, uses it. Students are neither — you are the affected person the rules protect.
Providers must run lifelong risk management (Art. 9), govern training and test data (Art. 10), write technical documentation (Art. 11), build in logging (Art. 12), transparency (Art. 13) and human oversight (Art. 14), and meet accuracy, robustness and cybersecurity requirements (Art. 15). Before use, the system needs a conformity assessment, CE marking and EU database registration.
Deployers — your university — must, under Article 26, use the system as instructed, assign competent human oversight with authority to intervene, keep input data relevant, monitor and report serious incidents, keep logs for at least six months, and inform people subject to a decision the system helped to make. Public universities will in most cases also need a fundamental rights impact assessment before first use (Article 27).
In practice, procurement changes first: a university buying a proctoring or admissions tool after 2027 will need the vendor's conformity paperwork in place before signing.
What you will actually notice on campus
Less than the headlines suggest: mostly procedural changes, arriving gradually through 2027.
More notices, in plainer language. Article 50 already requires deployers to tell you when you are interacting with an AI system, and Article 26 requires them to inform you when a high-risk system is involved in a decision about you. Expect short notices where today there is silence. Read them.
A human in the loop, for real. Human oversight becomes a named person with the authority to overrule the system. If an automated proctoring flag or an admissions score goes against you, a competent human must be able to look at the case and decide differently. If your appeal is answered with "the system decided", that is now the wrong answer by law.
Proctoring gets more careful, not more invasive. Expect clearer rules about what is recorded, how long recordings are kept, and who reviews a flag before it becomes an accusation.
AI-generated content gets labelled. Since 2 August 2026, AI-generated text, images, audio and video must be marked as machine-generated, and deepfakes disclosed as artificially generated. You will see more "AI-generated" labels on university communications.
Your right to an explanation — and how to use it
Article 86 is the part of the Act most useful to an individual student. It gives any person subject to a decision based on the output of a high-risk AI system listed in Annex III — except the research-related systems in point 2 — which produces legal effects or similarly significantly affects them, the right to obtain from the deployer "clear and meaningful explanations" of the AI system's role in the decision and its main elements.
It is a right to an explanation of your decision, not to the system's source code or training data. It applies where the AI's output significantly shaped the outcome — a rejected application, a failed grade, a proctoring accusation — not where the AI merely formatted a document. Under Article 86(3) it applies only where EU law does not already provide the same right — the GDPR's Article 22 already covers decisions based solely on automated processing.
How to use it: write to the office that made the decision and ask three things: which AI system was used and what it said about you; what role that played; and what the main elements of the decision were. If the university refuses, Article 85 lets any person complain to the relevant market surveillance authority. The general explainer puts this right in context see article #40: The EU AI Act Explained for Students: What Actually Applies to You in 2026.
What does not change in 2027
ChatGPT is not banned. Nothing in the 2027 rules stops you using a general-purpose chatbot for studying. What your course allows you to submit is a matter of university academic integrity rules, not the AI Act — and those rules are unchanged by 2 December 2027.
Students are not fined. Article 99 fines run to €35 million or 7% of turnover for breaching the prohibitions, and €15 million or 3% for breaching provider or deployer duties — but they fall on operators, not students. There is no fine in the Act for a student who uses AI in coursework; your university's disciplinary process is a separate matter.
The delay is not a repeal. The Digital Omnibus moved dates and added two bans; it did not cancel the high-risk regime. Anybody telling you the AI Act was "scrapped" or "delayed indefinitely" in 2026 is wrong: the dates are fixed in Regulation (EU) 2026/1744, in force since 27 July 2026.
What to do before December 2027
You do not need to do anything about the Act itself, but a little preparation helps. Read your university's AI policy, especially the exam section; keep drafts, prompts and version history for important work; ask which systems your university uses for admissions, grading and proctoring; and learn the appeal deadline and ombudsman route now — deadlines are short, seven days at some universities. Finally, check the date on anything you read about the Act: guides published before July 2026 still carry the old dates.
Frequently asked questions
Does the EU AI Act apply to me if I study in the EU but I am not an EU citizen?
Yes. The Act protects people in the EU regardless of nationality: the duties and your Article 86 right apply to you in the same way.
Will my university stop using AI proctoring in 2027?
Probably not — but it will have to use it properly. Proctoring is high-risk under Annex III, point 3(d), so from 2 December 2027 the system must meet the Chapter III requirements, with human oversight and logging. Some universities may switch to in-person exams instead.
Can I be failed by an AI system with no human involved?
Not lawfully. The AI Act requires effective human oversight, with a person able to intervene, and the GDPR already gives you rights over decisions based solely on automated processing.
I use ChatGPT for studying. Do I need to worry about 2 December 2027?
No. The 2027 date concerns the high-risk systems your university deploys. Your personal use of a chatbot is not a high-risk system, and the Act fines providers and deployers, not students — what matters is your course's own rules about declaring AI use.
The bottom line
2 December 2027 is not the day AI arrives at your university — it is the day the rules for the AI already there finally bite. Admissions, grading, placement and proctoring become high-risk in law, bringing risk management, human oversight, six months of logs, and your right to an explanation. The delay bought universities time; it did not change what they owe you. Read the notices, keep your drafts, know the appeal route.
Related guides
- The EU AI Act Explained for Students: What Actually Applies to You in 2026 (article #40)
- Your Exam Rights Under the EU AI Act: Proctoring, Emotion Recognition and the Right to an Explanation (article #47)
- How Universities Detect AI Writing in 2026 — and How to Stay Honest (article #19)
Facts checked against official pages in November 2026: Regulation (EU) 2024/1689 and Regulation (EU) 2026/1744 on EUR-Lex, and the Commission's AI Act pages. Secondary-source claims are flagged as such, including the position on AI detectors. General information, not legal advice.

