The experiment

Greece, ChatGPT enters schools: teacher training starts amid controversy

Next week, twenty high schools will train teachers on 'ChatGPT Edu', the educational version of the chatbot developed by OpenAI: this is the first step in the agreement between the Mitsotakis government and the Onassis Foundation to bring artificial intelligence into the heart of the Greek education system

by Angelica Migliorisi

22 November 2025

8' min read

Translated by AI

Versione italiana

Key points

"OpenAI for Greece"
"AI in Schools": what really happens in the twenty pilot schools
The promise of a domestic and controlled AI
The Mitsotakis vision: four axes for Greek AI
OpenAI Promises
Critical voices
Greece European laboratory, after Estonia

8' min read

Translated by AI

Versione italiana

Next week, in twenty Greek high schools, teachers will sit on the other side of the desk. For hours they will not have a class of teenagers in front of them, but a specialised version of ChatGPT, designed for educational institutions and introduced by a new agreement between the centre-right government of Kyriakos Mitsotakis and OpenAI.

It is the first concrete step in an experiment that can change the way teaching and learning is done in the country, and which many observers see as a test for the whole of Europe. For the executive it is "a new, great, exciting opportunity for the country". For some students and teachers, however, it is the beginning of a phase in which thinking will no longer be done by people, but by algorithms.

"OpenAI for Greece"

On 5 September, the Memorandum of Understanding 'OpenAI for Greece' was signed at the Hellenic Expo: on the one hand Prime Minister Kyriakos Mitsotakis and the relevant ministers, and on the other OpenAI, supported by the Onassis Foundation and Endeavor Greece. The stated aim is to bring high-quality artificial intelligence (AI) tools into secondary schools and accelerate innovation in the Greek startup ecosystem. On the educational front, the heart of the partnership is "ChatGPT Edu", a chatbot variant designed for schools and universities that offers access to the latest models, designed to be GDPR-compliant, running in a closed environment without advertisements, and able to enable centralised account management per institution and different roles for teachers and students. The Onassis Foundation is coordinating the implementation together with local partner "The Tipping Point in Education", while OpenAI is co-designing the training, providing technical support and sharing "best practices" for secure classroom use.

The "Greek AI Accelerator Program", an accelerator programme dedicated to Greek start-ups developing AI-based solutions, with access to technology credits, mentoring by OpenAI engineers and international visibility, has been launched. The ambition is to make Greece a technology hub capable of retaining the STEM graduates who all too often emigrate today and transform the country into a player in the so-called 'Intelligence Age'.

"AI in Schools": what really happens in the twenty pilot schools

The agreement is translated on the ground in the 'AI in Schools' programme, the pilot project that begins with the training of staff in 20 schools distributed throughout the territory, selected to represent different social and geographical contexts.These are mostly model and experimental high schools, to which are added six Onassis public schools, recently opened as a showcase of a public school 'strengthened' by private partnerships. The multi-stage timetable is structured as follows: 1) Between October and November 2025: intensive teacher training to learn about "ChatGPT Edu", understand its functions and learn how to integrate it into both teaching and administrative activities; 2) December 2025 to February 2026:first phase of "controlled" use by teachers, who begin to employ AI to prepare lessons, assessment materials, and exercises differentiated by level, as well as to lighten paperwork; 3) March to June 2026: starting specific training on how to get students to use "ChatGPT Edu"-initially the older ones-for research, projects, creative activities, and foreign language study, always under teacher supervision; 4) in the 2026-27 school year: Completion, with parallel use by teachers and students in schools that will have passed the first three phases with success.

On the practical side, teachers will have access to a panel that allows them to create 'customised GPTs' for specific subjects: an assistant for planning history lessons, one to generate maths quizzes at different levels of difficulty, and a virtual tutor for English that proposes dialogues and corrections in real time. Students, in turn, will be able to use AI to explore a topic in depth, generate ideas for a project, simulate conversations in the language they are studying, or break down a complex problem into simpler steps.

According to the Ministry of Education, Religious Affairs and Sports, thepurpose is not to replace the frontal lecture, but to accompany it with a 'customised digital tutor' to help meet the needs of children who learn at different rhythms. Therefore, the participation of teachers is voluntary.

The promise of domestic and controlled AI

One point on which the government, the Onassis Foundation and OpenAI are very insistent is the issue of data. "ChatGPT Edu" should operate in a "closed" environment, without using school data to re-train models, with information encryption and full compliance with GDPR, the European privacy regulation. The student licences - around 5,000 in the first phase - will be activated gradually and only when teachers have acquired sufficient skills to guide the children in an informed use.

From a formal point of view, the ministry activated a DPIA ('Data Protection Impact Assessment'), as required by European legislation, and involved the Data Protection Officer for an ongoing evaluation of the programme. The contracts with OpenAI, it is emphasised, expressly prohibit the sale or sharing of data collected in schools.

But the privacy issue is not only legal, it is also political and symbolic: a European state entrusts a crucial piece of its educational infrastructure to a large US platform. This is where the 'OpenAI for Greece' project ties in with the broader 'OpenAI for Countries' programme, which includes infrastructure partnerships such as 'Stargate Norway' for sovereign computing capabilities and agreements with the UK, Estonia, Ireland and Brazil. The idea is to build an ecosystem where AI is both a global service and a tool tailored to national goals.

Mitsotakis' vision: four axes for Greek AI

Prime Minister Kyriakos Mitsotakis often uses the image of the 'new great opportunity' to describe the trajectory of artificial intelligence in Greece. In the public presentation of the agreement with OpenAI, he punctuated four axes of intervention: uusing AI to improve the work of the state machine, starting in his own office, automating repetitive procedures and increasing the productivity of civil servants; toapplying AI to key areas such as healthcare - from diagnostics to the organisation of waiting lists - and education, where the state wants to move from a "de facto" situation (students already using ChatGPT informally) to an "organised and regulated" integration of generative tools; cbuilding an ecosystem of startups and innovative AI companies that choose to operate in Greecesupported by infrastructure, European funds and tax incentives, with the ambition of bringing Greek researchers currently abroad back to the country; rcreating a role for itself in the great international debate on AI ethics, drawing on the Greek philosophical and democratic tradition. Mitsotakis speaks explicitly of the possibility of making Greece a 'new Athens or new Delphi' of the world of artificial intelligence, a place where technology experts and social scientists meet to tackle the challenges of this technology.

Education occupies a central place in the premier's speech, also for very concrete reasons. The 'panelladikes', the national university entrance exams, are the most stressful step in Greek school life and feed a large market of private courses. The idea of a free and personalised 'personal digital teacher' for exam preparation is one of the strongest arguments used by the executive to win over families and public opinion.

The promises of OpenAI

On the OpenAI side, the tone is equally ambitious. Chris Lehane, Chief Global Affairs Officer, emphasises the symbolic continuity 'from Plato's Academy to Aristotle's Lyceum' to present Greece as a natural location for this experiment, recalling that almost 60 per cent of Greek ChatGPT users are under 35 years old. Sam Altman, ceo of OpenAI, pushes the comparison with the calculator: when it was introduced, many feared the end of mathematics teaching; in reality, by freeing it from mechanical exercises, it allowed people to raise the bar and tackle more advanced problems.

According to Altman, generative AI will be "a calculator for words", a tool that, if intelligently integrated into the educational pathway, will make students more capable and productive than today's adults. At the same time, he acknowledges that there are 'less positive effects' to mitigate, from misinformation to psychological dependence, and identifies collaboration with governments and 'smart guardrails' as the way to contain them.

And it is here that an issue comes into play that will weigh heavily in the Greek debate:OpenAI is fresh from a lawsuit brought by a US family that attributes the suicide of its 16-year-old son to a psychological addiction to ChatGPT, and has just announced new 'parental control' tools to limit access and monitor use among minors. Bringing the same tool into schools in a structural way therefore requires a leap in transparency on how it is used, what the risks are, what the reporting and blocking mechanisms are.

Critical voices

If the institutional and corporate narrative is punctuated with words like 'opportunity', 'innovation', 'new educational chapter', very different is the tone emerging from a section of professors and students. Among high-school teachers, gathered in the OLME federation representing some 85,000 faculty members, some of whom were interviewed by the Guardian, there is a fear of a gradual 'disintermediation' of their role, up to the horizon - for now still remote - of teacherless classrooms, with platforms managing much of the learning journey.

And it is not just about corporate defence: many point out that introducing AI into a system dominated by memorisation and standardised test preparation risks reinforcing the very thing that official rhetoric says it wants to overcome. All these concerns are amplified by the fact that Greece is on the verge of becoming the first European country to block access to social media for under-15s, precisely to counter the impact of platforms and digital addiction on mental health.

Then there is a more material criticism: how can aschool system that historically devotes less than 5% of the national budget to education, with schools where in winter the heating runs for an hour a day and the electrical sockets do not always hold up, suddenly be ready for the 'school of the future' powered by artificial intelligence? Some teachers argue that the priorities should be infrastructure, salaries, reducing the number of students per class, and only after a reflection on the use of advanced tools.

Greece European laboratory, after Estonia

Yet, Greece is not the first country to experiment with 'ChatGPT Edu' on a national scale. Estonia announced this year the 'AI Leap' programme, which brings an educational version of ChatGPT to all secondary schools in the country, with 20,000 students and 3,000 teachers involved in the first phase. There, AI is presented as a tool to strengthen critical thinking, not to replace teachers. Athens follows suit and, with the 5 September agreement and the 'AI in Schools' programme, is bidding to become one of the first cases of structural integration of generative AI in the curriculum of a large public education system in the Mediterranean area. In parallel, OpenAI has launched similar initiatives with other governments, from Ireland to Brazil, while in several European countries education ministries are drawing up guidelines on the use of AI in the classroom, often straddling the line between promotion and caution.

In the meantime, in the classrooms of the twenty pilot schools, between the teacher who discovers a digital assistant capable of preparing a lesson in a few minutes and the student who fears being 'overtaken' by a soulless machine, a game is already being played that is not just about Athens. It is the first concrete test of how Europe intends to educate its citizens in the age of artificial intelligence.