Who gets to build the AI-powered classroom? Greece’s AI education experiment and the New Cold War by Dr Aristea Fotopoulou
When Greece announced it would deploy ChatGPT Edu across selected secondary schools in October 2025, the news generated limited traction in a national media landscape, which is alarmingly dominated by political/economic scandals and femicides. The Memorandum of Understanding, signed at the Hellenic Expo in Thessaloniki in September 2025, positions Greece among the first countries worldwide to pilot large-scale use of OpenAI’s technology in public education (Estonia was one of the first countries to form a partnership with OpenAI). The Onassis Foundation, the country’s most prestigious philanthropic institution, was chosen to lead implementation in its selective public schools (‘πρότυπα’). The framing is optimistic: personalised learning, AI-ready graduates, and a nation preparing its youth for a transformed labour market.
But the policy significance of this initiative extends beyond Greece’s education system as it comes at a time of intense geopolitical changes with immediate consequences to technological dependency. Two weeks ago, the US ordered Anthropic suspended access to its frontier technologies (the most advanced Fable 5 and Mythos 5 models) for all foreign nationals, citing national security concerns. Major financial institutions, including JPMorgan and Goldman Sachs, immediately extended the restrictions not only to China, but also to Hong Kong-based staff. Although these restrictions have now been lifted and frontier models are available to selected companies chosen by the US Commerce Department’s Bureau of Industry and Security, such measures inevitably impact directly on the research capacity of European universities and policy makers. Of course, the competition between the US and China in the field of AI, what has been referred to by commentators as the new Cold War, involves hardware as well, including semiconductors, data centres and chip fabrication lines. Like other regions, the European Commission immediately voiced opposition to the restriction measures, and urged for European technological sovereignty in AI, which is a priority for the region and further established in the forthcoming Cloud and AI Development Act (CAIDA). Within this context of sanctions and export controls, where does AI education policy in European countries stand?
European countries are very uneven in terms of AI infrastructure, policy priorities and implementation stage. For prime movers, like Germany and France, who have a robust digital infrastructure and an advanced policy mechanism that allows stakeholders to participate in policy-making, the US restrictions just signal a more urgent need for a European AI sovereignty. But this is already a priority in AI policy, which aims to accelerate investment in European AI champions such as Mistral, and strengthened commitment to the EU AI Act as a framework for managing risk on European terms.
When it comes to Greece, and other peripheral EU states like Poland or Portugal, US restrictions that limit access to advanced GPUs create an infrastructure bottleneck that potentially impacts the development and deployment of advanced educational AI tools. Here, teacher training and digital infrastructure are at an embryonic stage, which already stifles implementation of the Digital Education Action Plan (2021-2027), a highly ambitious educational EU policy aimed to ‘future proof Europe’s education systems’ at Member State level.
My forthcoming contribution to the Critical AI Policy Fellowship Report traces how Greece arrived at this policy juncture. The key driver has been the National Recovery and Resilience Plan, Greece 2.0, launched in 2021, framing digital transformation as a historical mission. However, deep national anxieties about the emigration of highly qualified young professionals (the so-called Greek ‘brain drain’) is possibly the main reason for latching on to education as the frontier where workforce retainment is fought for Greece – effectively hoping to turn education into a pipeline for digital skills and entrepreneurship. Indeed, the 2026 OpenAI partnership has been framed as both an educational intervention and a ‘Greek AI Startup Accelerator’, two years after a high-level Advisory Committee on AI in Education was formed to produce an ethical framework for AI integration. The initiative operates within a public education system that remains profoundly under-resourced, with structural challenges ranging from teacher shortages to deteriorating physical infrastructure. Teachers, parents, and students remain largely uninformed and absent from the consultation, a point raised by 22 civil society organisations and experts in March 2026. The problematic framing raises questions of policy coherence: if the intended beneficiary is the student, why is the student positioned primarily as a future entrepreneur? If the initiative is designed to serve national economic development, what governance process determined that the secondary classroom is the appropriate starting point, and what risk assessment informed that decision?
The capacity to manage AI risk for students and educators is questionable for Greece, and it comes at a point of yet another further policy tension: the EU AI Omnibus (May 2026), a simplification of the AI Act for Europe, introduced in order to enhance competitiveness and to streamline administrative load. The package has been criticised by educators’ unions across Europe, because it removes mandatory AI literacy obligations and delays the application of high-risk classification to educational contexts. The Omnibus extends the implementation deadline for these Annex III tools to December 2, 2027 (instead of August 2026) – and this is software used for student admissions, grading, exams and monitoring student behaviour. Although the Greek state has clarified that the use of AI will not apply to student assessment, so seemingly exempt from legally binding conformity assessments, the permissiveness or, in fact, temporary deregulation presented by the AI Omnibus while OpenAI rolls out in Greek secondary classrooms is a danger zone for countries like Greece.
To wrap up, for a country that has chosen to become dependent on a US proprietary LLM rather than invest in or acquire the EU funds to develop domestic AI capacity (if not sovereignty), the implications of the Anthropic restrictions are severe because they signal the abrupt shifting terms of access imposed by the US. Even though there is currently no indication that there will be a ban on ChatGPT Edu, the Greek educational system is facing yet an additional strategic vulnerability, with September 2027 quite far off and the memorandum of understanding a non-legally binding agreement. The cut lifeline for researchers to advanced software adds a huge disadvantage to countries that basically rely on software, since they lack expertise and AI research capacity, and it creates a further achievement gap. Therefore, the critical policy questions here extend beyond pedagogical applications onto strategic interests of actors larger than those of peripheral EU state members. If anything, the Greek case shows how the shifting geopolitics of AI are about to hit the classroom.
Dr. Aristea Fotopoulou is a 2026 Critical AI Policy Fellow researching the governance of artificial intelligence in Europe, with emphasis in education. Her work examines how AI policy intersects with sovereignty, ethics, equity and democratic accountability. Beyond the Fellowship, she has been advancing creative interventions for developing critical AI and data literacies, while her wider research focus is on the social aspects of emerging technologies. She blogs as PixelScholar of Substack https://pixelscholar.substack.com/
LinkedIn: https://www.linkedin.com/in/aristea-fotopoulou-24903819/
