Tartu ethics center: Estonia's AI Leap education program needs careful management of risks
A working group at the University of Tartu Centre for Ethics has published an analysis of risks tied to Estonia's AI Leap education program and its AI-based learning application. The report warns that the greatest dangers come not from the technology itself but from how it is used and the policy choices made around it.
In mid-September, a working group at the University of Tartu Centre for Ethics published an analysis identifying 12 risks and related recommendations concerning Estonia's AI Leap (TI-Hüpe) education innovation program and its AI-based learning application.
The authors stress that introducing such an application in schools is not merely a technical change — it affects learning quality, teachers' work, educational equality, the sustainability of the Estonian language and culture, and the broader value system underlying education.
Risks go beyond technology
The 12 identified risks fall into three categories: pedagogical, societal, and infrastructure/governance. This shows the challenges extend beyond cybersecurity to questions about the goals of learning, the reliability of assessment, the role of teachers, and the future of Estonian as a language of education.
An early-2024 study found that 90 percent of upper secondary students in Estonia had already used AI, mostly free versions of commercial tools, to complete schoolwork faster. The authors warn that a quickly produced result does not necessarily mean real learning has occurred — a student may submit flawless work without being able to explain its reasoning.
Teachers must keep their leading role
The Socratic learning application being developed through a collaboration between OpenAI and Estonian researchers differs from an ordinary chatbot, aiming to guide students toward answers through questions and hints rather than giving them directly. However, the authors caution that AI must not become an invisible, parallel teaching authority — teachers must retain pedagogical leadership.
A fall 2024 survey found that 47 percent of teachers had not yet used AI tools in their work, underscoring the need for further training and support.
The report also warns that international large language models may not draw sufficiently on Estonian-language educational materials, risking a situation where Estonian becomes merely a user-interface language while more complex content is drawn from English-language sources. The authors recommend treating the risk analysis as an ongoing management tool for the program rather than a one-time document.
