The European Union’s Artificial Intelligence (AI) Act is a landmark regulation setting global standards for AI systems. Understanding its implications is crucial, especially within the sensitive sector of education.
Whether you’re an educator, administrator, or EdTech developer using or creating AI tools for learning, assessment, or administration, navigating the compliance requirements based on potential risks is essential.
EU AI Act Risk Assessment Tool (Education Focus)
This simplified tool helps assess the potential risk category of an AI system used in education under the EU AI Act. Answer the following questions:
(Examples: manipulative techniques causing harm, exploiting vulnerabilities of specific groups, social scoring by public authorities, certain uses of real-time remote biometric identification in public spaces).
- Determining access or admission to educational/vocational institutions?
- OR Evaluating learning outcomes, assessing test participants for admission, or assigning individuals to educational programs?
(These are listed as high-risk use cases in Annex III, point 5 of the EU AI Act).
- It performs a narrow procedural task.
- OR It’s intended solely to improve the result of a previously completed human activity (e.g., grammar check on an already written essay).
- OR It detects decision-making patterns or deviations from prior patterns without directly influencing an assessment or decision about a person (e.g., plagiarism detection *after* grading).
(If the AI *does* materially influence the outcome, e.g., automated grading that determines pass/fail, it likely does *not* meet these exceptions).
- Direct interaction with humans where the human needs to know they are interacting with AI (e.g., chatbots, virtual tutors)?
- Emotion recognition or biometric categorization?
- Generation or manipulation of image, audio, or video content constituting a ‘deep fake’?
(These generally trigger transparency obligations under Article 50).
Outcome: Unacceptable Risk (Prohibited)
AI systems falling under this category are generally prohibited under the EU AI Act as they contravene Union values or fundamental rights.
Outcome: High-Risk
This AI system is likely classified as ‘High-Risk’ under the EU AI Act for educational use cases. It must comply with stringent requirements regarding risk management, data governance, technical documentation, transparency, human oversight, accuracy, robustness, and cybersecurity before being placed on the market or put into service.
Outcome: Limited Risk
This AI system likely falls into the ‘Limited Risk’ category. Specific transparency obligations apply. Users must typically be informed that they are interacting with an AI system, or that content (like deepfakes) has been artificially generated or manipulated.
Outcome: Minimal Risk
Based on these answers, the AI system likely falls into the ‘Minimal Risk’ category. The EU AI Act does not impose mandatory obligations for these systems, although voluntary codes of conduct are encouraged to ensure trustworthiness.
Disclaimer: This tool provides a simplified indication based on the EU AI Act and focuses on common educational scenarios. It is not exhaustive and does not constitute legal advice. Consult the official text of the Regulation and seek legal counsel for definitive compliance assessments.
How the EU AI Act classifies AI tools used in education
The EU AI Act sets four risk categories: prohibited, high-risk, limited-risk (transparency-only), and minimal. Classification depends on what the AI system is used for in your institution, not on the brand of tool. The same general-purpose chatbot can be minimal in one deployment and high-risk in another, depending on whether it determines an outcome that affects a student.
Function-by-function classification
The table below maps the AI use cases that come up most often in schools, colleges, and universities to the most likely classification under the Act, with the specific provision that drives it. This is a working reference, not legal advice.
| What the AI is used for | Most likely classification | Provision in the Act |
|---|---|---|
| Emotion recognition of students or staff | Prohibited | Article 5(1)(f): emotion recognition is prohibited in education and workplace contexts, with narrow safety and medical exceptions |
| Social scoring of students | Prohibited | Article 5(1)(c) |
| Determining admission to a school, college, or programme | High-risk | Annex III, point 5(a) |
| Automated grading or assessment that decides whether a student passes, fails, advances, or is sorted into a level | High-risk | Annex III, point 5(b) and 5(c) |
| AI proctoring or monitoring during tests to detect prohibited behaviour | High-risk | Annex III, point 5(d) |
| AI that personalises the learning path in a way that materially steers what a student is taught next | High-risk | Annex III, point 5(b), where the system materially influences learning outcomes |
| AI tutor or chatbot interacting with students directly, with no graded outcome | Limited risk | Article 50(1): the user must be informed they are interacting with AI |
| AI generating study materials, audio, or images shown to students | Limited risk | Article 50(2): synthetic content must be marked as AI-generated |
| Plagiarism detection used after a teacher has independently graded the work | Likely exempt | Article 6(3) exception: detects patterns without influencing the prior assessment |
| Plagiarism detection that determines pass or fail before any teacher review | High-risk | The Article 6(3) exception does not apply when the AI materially influences the outcome under Annex III, point 5(b) |
| AI-assisted lesson planning, where the teacher reviews and edits the output | Minimal risk | Not in Annex III; the Article 6(3)(d) preparatory-task framing also applies |
| AI summarising or correcting work that has already been graded | Likely exempt | Article 6(3)(b): improves the result of a completed human activity |
| Administrative AI such as timetable optimisation or room booking | Minimal risk | Not in Annex III |
The most common misclassifications
- “It’s just a chatbot, so it’s minimal risk.” Limited risk, not minimal. Article 50(1) requires you to tell users they are interacting with AI.
- “We’re using it for grading but the teacher reviews it first.” If the teacher’s review is rubber-stamping rather than independently arrived at, the AI is materially influencing the outcome and Annex III, point 5(b) still applies. The Article 6(3)(d) exception requires the AI to perform a preparatory task, not the assessment itself.
- “Emotion recognition just helps us understand engagement.” Article 5(1)(f) prohibits emotion recognition in education with very narrow safety and medical exceptions. Repackaging it as “engagement analytics” does not change the classification.
- “We disabled the high-risk feature, so we’re fine.” Classification is by intended use. If the feature exists and could be used as intended, the obligations attach unless the deployer can demonstrate the use case cannot occur in their deployment.
If your AI is classified as high-risk: what comes next
A high-risk classification is not a ban. It triggers obligations, the most relevant for education being:
- A documented risk management system, including bias mitigation
- Data governance covering training data quality and representativeness
- Technical documentation of the system’s intended purpose, design, and limitations
- Logging sufficient to trace the system’s decisions
- Transparency to deployers and end users
- Human oversight built into the deployment, not just claimed in policy
- Accuracy, robustness, and cybersecurity testing appropriate to the risk
The deployer (the school, college, or university) carries obligations distinct from the provider (the AI vendor). A school using a third-party high-risk AI system still has duties around staff competence, monitoring, and incident reporting. Talk to Lumination if you need help structuring this for your institution.
Important Note
Please remember that this tool provides a simplified, preliminary assessment focused on educational scenarios under the EU AI Act. It is not a substitute for legal advice or a full compliance audit. Always consult the official text of the Regulation and seek professional legal counsel for definitive assessments and guidance specific to your situation.
Source: EU AI Act
Related tools from Lumination
- AI Literacy Self-Assessment for Educators – gauge your readiness across the EU AI Literacy and UNESCO competency frameworks.
- AI Policy Generator for Educational Institutions – build a first-draft AI usage policy for your school or university.
- AI Consulting for Education – work with Lumination on EU AI Act compliance, governance, and rollout.