This page provides an overview of an independent analysis by the Centre for Ethics, University of Tartu, examining the risks associated with the AI Leap educational programme and its AI-based learning application. The purpose of the analysis is to help identify potential risks at an early stage and strengthen the educational programme’s measures so that the use of AI supports learning and teaching and the values of Estonian education.
Risk Assessment
Why is an independent risk assessment needed?
A risk assessment helps to evaluate the programme’s potential impacts, identify blind spots at an early stage, strengthen existing mitigation measures and make informed choices while the programme is still being developed.
The analysis considers the AI Leap AI-based learning application as part of the broader education system, encompassing not only technology but also teaching practices, school culture, human relationships, data use, technology partners and public responsibility.
The risk scenarios do not mean that the harms described will necessarily materialise. Mitigating the risks requires cooperation between the AI Leap Foundation, schools, the state, research institutions, technology partners and the home. As AI and its uses are developing rapidly, the risks, the effectiveness of mitigation measures and the programme’s impact must all be reassessed regularly.
The central conclusion of the risk assessment is that AI can significantly strengthen Estonian education, but only if educational goals guide the use of technology, not the other way around.
Authors of the analysis
Read the report
The project “Ethical, Educational, Social, Legal and Technological Risk Assessment and Value Analysis of the AI Leap Learning Application and Educational Programme” was carried out by a working group of the Centre for Ethics, University of Tartu, and commissioned by the AI Leap Foundation. The work took place from 22 January 2026 to 30 June 2026.
The work was led by Margit Sutrop, Head of the Centre for Ethics and Professor of Practical Philosophy at the University of Tartu, and Katrin Laas-Mikko, AI ethics specialist at the Centre for Ethics (University of Tartu, SK ID Solutions AS).
The expert team also included Dan Bogdanov, Research Director at Cybernetica AS; Liina Kamm, Senior Researcher; Henrik Trasberg, Legal and Policy Officer at the European Commission AI Office; Andra Siibak, Professor of Media Studies at the University of Tartu.
How was the risk assessment conducted?
1. Programme mapping
The working group analysed the programme, the learning application and related documents, and interviewed key people involved in AI Leap. The analysis examined the technology, teaching practices, school culture, data use and governance.
2. Risk assessment
The experts assessed the possible causes and consequences of each risk and its impact on educational goals and ethical values.
3. Designing mitigation measures
For each risk, the analysis described existing and additional mitigation measures and identified the parties responsible for implementing them.
Three areas of risk
The analysis identifies 12 potential risks that must be considered when using AI in education. They fall into three areas.
Educational risks
How might AI affect the aims of learning, the reliability of assessment, the teacher’s role and schools’ capacity for change?
-
Goals of Estonian education
The AI Leap programme and the AI-based learning application may not sufficiently support the agreed goals of Estonian education, general competences and the student’s holistic development.
-
Academic integrity
Use of the AI-based learning application may undermine the reliability of assessment and the culture of honest learning.
-
Learning relationship
Use of the AI-based learning application may weaken the teacher’s pedagogical leadership and the teacher–student relationship.
-
Teachers’ capacity for change
Use of the AI-based learning application may increase teachers’ perceived workload and change fatigue, and reduce their capacity to implement change.
-
School responsibility
Schools may have limited means and control to fulfil the legal obligations associated with the use of the AI-based learning application and to protect the rights of students and teachers.
Societal risks
How might AI affect the Estonian language and culture, our shared values and human interaction?
-
Estonian culture
The AI-based learning application may not be sufficiently grounded in the Estonian curriculum, Estonian-language learning resources and the local cultural context, and may therefore threaten the sustainability of the Estonian language and culture.
-
Value drift
An AI-based learning application controlled by an external service provider may mediate values that conflict with the fundamental values of the Estonian state.
-
Alienation
Widespread use of AI-based learning applications may foster the anthropomorphisation of AI and students’ disengagement from human and social interaction.
Technology and governance risks
How can data protection, operational continuity, cost control and Estonia’s freedom to make its own decisions be ensured?
-
Data breach
Data entered by students and teachers into the learning application, or generated through its use, may become accessible to unauthorised persons or be used contrary to the rights and interests of the persons concerned.
-
Cost growth
A price increase by a service or infrastructure provider supporting the AI-based learning application may make the solution burdensome for the state budget.
-
Operational continuity
A service supporting the AI-based learning application may fail and thereby impede education.
-
Domestic solution
An AI-based learning application built using Estonia’s own competences and infrastructure may be lower in quality and security than a solution based on a foreign cloud service.
Main recommendations of the report
1
Evaluate AI Leap against the goals of education.
Use of the learning application should be evaluated according to whether it supports the learner’s holistic development, deep learning, teacher professionalism, and the values and goals of Estonian education.
2
Treat AI Leap as a comprehensive educational innovation.
A Socratic learning application alone will not ensure deep learning. Teaching practices, assessment, learning materials, teacher education, school culture and governance must develop alongside the technology.
3
Ensure clear responsibilities and sustainable national governance.
AI Leap’s technological and organisational system should be treated as public educational infrastructure that is developed, funded, evaluated and updated throughout its life cycle. Changes should be implemented gradually and on a research-informed basis, and their impacts should be evaluated continuously.
4
Strengthen teachers and learners, rather than replacing them.
Teachers must retain an overview of how learners learn, think and use AI so that they can support their development. AI must not undermine the learning relationship between teacher and student or encourage disengagement from learning and human interaction.
5
Safeguard Estonia’s educational self-determination.
The learning application must be grounded in the Estonian curriculum, high-quality Estonian-language learning resources and Estonia’s cultural context. The state must retain control over the core functions of education, data and technological solutions.
6
Systematically involve universities and research institutions.
The use of AI in education requires educational, didactic, psychological, philosophical, ethical and legal expertise. Researchers must be involved in designing solutions, teacher education, the development of learning materials, impact assessment and continuous risk monitoring.
"Artificial Intelligence may provide answers, but schools must teach students how to determine whether an answer is correct and what to do with it. AI must support, not replace, the human learning relationship. AI Leap gives us an opportunity to rethink what Estonian schools teach, how they teach it and for what purpose."
Margit Sutrop
Head of the Centre for Ethics, University of Tartu, and Head of the Risk Assessment Working Group
Risks and actions already under way
The central conclusion of the risk assessment is that AI can significantly strengthen Estonian education, but only if educational goals guide the use of technology, not the other way around. Explore the main risks and the actions already being taken to mitigate them.
1. Does the use of AI in teaching support the goals of Estonian education?
Risk
The AI Leap programme and the AI-based learning application may not sufficiently support the agreed goals of Estonian education, general competences and students’ holistic development.
Existing measures
The pedagogical principles of the learning application have been developed in cooperation with Estonian education researchers, and a long-term study has been commissioned to assess its impact. The programme emphasises meaningful learning and the development of critical AI literacy.
2. Will assessment remain reliable?
Risk
The use of artificial intelligence may undermine the reliability of assessment and the culture of honest learning.
Existing measures
AI Leap supports schools in establishing common principles for AI use and teachers in reconsidering assessment so that greater attention is given to the learning process. The learning application supports learning; it is not used to assess students.
3. Does the teacher remain in charge of the learning process?
Risk
The use of the learning application may weaken the teacher’s pedagogical leadership and the teacher–student relationship.
Existing measures
The programme primarily consists of training and seminars that develop teachers’ knowledge and practical skills in using AI. The programme explicitly places teachers at the centre: the learning application supports a teacher-led learning process and does not replace teachers’ pedagogical judgement or feedback.
4. Can teachers implement the changes?
Risk
The use of the learning application may increase teachers’ perceived workload and change fatigue and reduce their capacity to implement change.
Existing measures
In addition to training, AI Leap supports the establishment of learning circles, collaborative professional learning and the sharing of experience in schools, so that AI use reflects schools’ actual needs. Subject-specific sample materials are also being developed, and good practices are being collected and shared.
5. Can schools govern the use of artificial intelligence?
Risk
Schools may have limited means and control to fulfil the legal obligations associated with using the learning application and to protect the rights of students and teachers.
Existing measures
The roles of the parties, terms of use and data-processing conditions are set out in agreements between the AI Leap Foundation, technology providers and schools. Schools do not have to address these matters alone. The Foundation has conducted a data protection impact assessment and provides schools with training, guidance materials and technical support.
6. Does the AI-based learning application support the Estonian language and culture?
Risk
The learning application may not be sufficiently grounded in the Estonian curriculum, Estonian-language learning resources and cultural context, and may thereby threaten the sustainability of the Estonian language and culture.
Existing measures
Cooperation with the Institute of the Estonian Language includes evaluating the learning application’s conformity with Estonian-language and cultural criteria and improving the availability of Estonian-language material to large language models.
7. Does AI mediate values consistent with shared societal values in Estonia?
Risk
An AI-based learning application controlled by an external service provider may mediate values that conflict with the fundamental values of the Estonian state.
Existing measures
Developing critical AI literacy is one of the programme’s central aims. Training materials and teacher learning circles emphasise that AI outputs are not value-neutral and should be critically evaluated together with the teacher.
8. Does AI change the relationship between humans and technology and what it means to be human?
Risk
Widespread use of AI-based learning applications may foster the anthropomorphisation of AI and students’ disengagement from human social interaction.
Existing measures
The learning application is intended primarily for teaching and learning, and training emphasises that it is a tool with limited capabilities. Many schools have internal rules governing the use of learning tools and technology.
9. Are students’ and teachers’ data protected?
Risk
Data entered by students and teachers into the learning application, or generated through its use, may become accessible to unauthorised persons or be used contrary to the rights and interests of the persons concerned.
Existing measures
All parties involved in the data-processing chain apply data-protection and cybersecurity measures, including access controls. Data-processing conditions and the responsibilities of the parties are set out in agreements. The user interface warns users not to enter sensitive information or personal data.
10. Will the AI-based learning application remain affordable for the state?
Risk
A price increase by a service or infrastructure provider supporting the AI-based learning application may make the solution burdensome for the state budget.
Existing measures
The initial technical implementation of the learning application has been kept relatively simple, and provider-specific solutions that would make switching providers expensive or difficult have been avoided.
11. Will education continue to function if the AI-based learning application becomes unavailable?
Risk
A disruption to a service supporting the AI-based learning application may interfere with teaching and impede education.
Existing measures
Use of the learning application is not a compulsory or critical part of teaching and learning. Education can continue if the service becomes unavailable. AI Leap also provides users with technical support when necessary.
12. Can Estonia create a trustworthy AI-based learning application?
Risk
An AI-based learning application built using Estonia’s own competences and infrastructure may be lower in quality or security than a solution based on a foreign cloud service.
Existing measures
The solutions currently in use rely on services provided by the world’s leading large language model developers, whose quality and security have been subject to international audits.