EuroMedAI programme · since April 2026
Mediterranean AI Safety Fundamentals
A monthly seminar where EuroMedAI members work through the core of AI safety—how advanced AI systems can fail, why the stakes could be catastrophic, and what researchers and governments are doing about it—and apply it to the Mediterranean.
See what we studyThe programme
Serious about safety, grounded in the region.
Mediterranean AI Safety Fundamentals gives EuroMedAI’s researchers, policy specialists and practitioners a shared, rigorous grounding in AI safety. Each month, members read foundational work from the field and debate it in a session led by EuroMedAI’s leadership, connecting the global research agenda to the institutions and policy choices of the region.
- Started
- April 2026
- Rhythm
- One session a month
- Format
- Readings and facilitated debate
- Participants
- EuroMedAI members
Why the Mediterranean
Frontier AI risks cross borders. So must the expertise to manage them.
The most capable AI systems are built in a handful of places, but their consequences will reach every society. Misuse, accidents and loss of control are not problems any one country can contain on its own.
The Mediterranean joins the European Union, home to the world’s first comprehensive AI law, with North African and Middle Eastern countries that are largely absent from today’s safety debates. A network fluent in the technical case for safety can carry it across both shores, and make sure the region helps write the rules rather than simply receiving them.
What we study
Six themes, from first principles to regional policy.
The readings follow the core questions of the AI safety field and close on what they mean for the Mediterranean.
- 01
Capabilities and trajectories
How frontier models are built and scaled, what they can already do, and what current trends imply for the coming decade.
- 02
The alignment problem
Why it is hard to make capable systems reliably pursue the goals we intend, from specification gaming to goal misgeneralisation and deception.
- 03
Pathways to catastrophe
Misuse, accidents, loss of control and structural risks, and how researchers reason about their likelihood and severity.
- 04
Technical safety research
Interpretability, scalable oversight, evaluations and red-teaming: what each research agenda is trying to solve, and where it falls short.
- 05
Governing frontier AI
Safety frameworks, compute governance, the EU AI Act’s obligations for general-purpose models, and the case for international coordination.
- 06
The Mediterranean stake
What these risks mean for countries that mostly deploy rather than build frontier AI, and where regional institutions have real leverage.
Theory of change
From the reading room to decisions that reduce risk.
The seminar is the first stage of EuroMedAI’s AI safety work. Each stage builds the capacity the next one needs.
01 · Learn
Shared foundations
Members gain fluency in the core arguments, threat models and research agendas of AI safety.
02 · Reason
Sharper judgement
They learn to separate catastrophic-risk questions from adjacent tech-policy debates, and to weigh evidence and uncertainty with rigour.
03 · Localise
Regional relevance
They map where Mediterranean institutions, infrastructure and policy choices intersect with risks from advanced AI.
04 · Act
Research and policy work
The strongest questions feed EuroMedAI’s research, policy engagement and talent development, and collaboration with organisations working directly on advanced-AI safety.
05 · Goal
Lower catastrophic risk
Better-informed decisions and stronger coordination across the region, helping reduce the likelihood and severity of catastrophic outcomes from advanced AI.
Who leads it
Convened by EuroMedAI’s leadership.
The seminar is convened by Dr. Alejandro Tlaie Boria and co-led by EuroMedAI’s two other chairs, combining technical AI safety, governance research and regional policy experience.
Meet the full team
Programme convener
Dr. Alejandro Tlaie Boria
AI Policy Advisor and Talos Fellow, SaferAI · Co-Chair for Europe, EuroMedAI
Co-lead
Ermelinda Kanushi
AI Governance Researcher, University of Technology Nuremberg · Co-Chair for the Balkans, EuroMedAI
Co-lead
Mohamed Alouane
Founder and Chair, EuroMedAI
EuroMedAI · AI safety
Work with us on AI safety.
Sessions are reserved for EuroMedAI members. Researchers and organisations working on advanced-AI safety can contact us to lead a session, co-develop research or bring their work to Mediterranean audiences.