The Unexpected AI Unicorn in Europe: Albania

This article is an opinion piece by a current student or alumnus/alumna of the College of Europe, featured in our monthly newsletter. The views expressed are those of the author and do not necessarily reflect the opinions or positions of the College of Europe. Responsibility for the content lies solely with the author.

 

By Amy KOKALARI & Timothy YEUNG

“Albania has finally entered the age of Artificial Intelligence!" The news broke out in Albania in December 2023 when Prime Minister Edi Rama announced that the government would use ChatGPT to accelerate its EU accession process. This partnership clearly suggests that joining the EU is considered solely as a technical process.

The Albanian government has signed a formal agreement with Mira Murati, the executive director of OpenAI, who also happens to be of Albanian origins, indicating their shared commitment to this AI-driven project. Following this arrangement, professionals from OpenAI and the National Agency for Information Society (AKSHI) have joined forces to form a collaborative team. This partnership aims to do more than just translate EU legislation to Albanian, but it will also analyse more than 280,000 pages of EU legislation and identify areas where the Albanian legal system must be changed to bring it into compliance with the European acquis communautaire. Moreover, the AI tool will evaluate the expected effects of these changes, providing a detailed overview of the upcoming modifications. According to Prime Minister Rama, this new strategy not only eliminates the requirement for "a battalion of lawyers and an army of translators", but it also streamlines the involvement of multiple experts, resulting in a significant reduction in the costs and timeline related to this complex process. In this case, using ChatGPT turns out to be a practical and affordable approach that might accelerate the process of bringing Albanian legal systems into compliance with EU regulations.

This is a small part of Prime Minister Rama's larger plan, which aims to utilize the chatbot for various local and national administrative tasks. The main goal is to combat corruption, a critical issue impacting Albania and other Balkan countries seeking EU membership. In 2024, Rama doubled down on his AI ambitions. Building on the initial ChatGPT initiative, he announced his plans to use AI in Albania’s public procurement system - framing it as a way to “neutralize” the procurement process and fight corruption. The timing is no coincidence. With national elections looming and SPAK’s corruption investigations closing in on high-level political figures, Rama is eager to implement his plan. And he’s not stopping at ChatGPT. In a recent plenary session, the Prime Minister announced a partnership with Microsoft to develop the first fully AI-powered, contactless procurement platform in Europe - a model he claims will set the standard for the rest of the continent.

With the strategic objective of advancing the country's quest for EU membership, the innovative initiative centers on the integration and regular use of AI devices, particularly ChatGPT, in the country’s EU accession process. This technological advance, according to Prime Minister Edi Rama, will be the long-awaited tipping moment that brings Albania into the EU and welcomes a new period of development and integration.

Cheaper, smarter, faster. At least in theory.

However, the Albanian Prime Minister is showing an excess of optimism, and there is a need for a reality check. As EU Ambassador Silvio Gonzato pointed out: “Technology is definitely a help... but it will still need a lot of human factor.”

And that’s the key. No matter how fast ChatGPT processes documents or how a Microsoft platform automates procurement, accession is not a mere technical process. It is political. It is deeply human. And by nature, it is often slow, difficult, and full of compromise.

Albania’s AI facade: masking deep structural challenges

Albania may become a leader in the efficient incorporation of AI into European expansion processes. It can establish itself as a digital hub that will eventually inspire medium- and large-sized countries. However, these visionary goals must be balanced with various considerations that appear to have been overlooked by the Albanian government.

It is important to highlight the need to closely examine ChatGPT's interpretation of the legislation. Assessing compliance with the legislations of a candidate state can’t be detached from political factors, such as the political will, power dynamics, and institutional structures that weigh in on accession. The process is not just paperworkbut it’s layered and deeply political.

Even though AI mimics certain activities performed by human intelligence, it is fundamental to emphasise that AI and human intelligence are different. Moreover, currently generative AI has not gained sufficient trust from the public. For example, just a couple of days agoa Norwegian man filed a complaint after the AI falsely claimed he had murdered his children, demonstrating how AI chatbots responses can result in false information and wild assertions. The trust issue is certainly linked to the potential production of inaccurate information. Consequently, there exists a risk that blind reliance on these tools may inadvertently extend accession processes, given the intricate nature of the European acquis communautaire and the possibility of officials uncritically adhering to ChatGPT's evaluation.

When it comes to trust, in Albania also public trust in public institutions remains low, and Transparency International still ranks Albania among the most corrupt countries in Europe. If misused, AI risks becoming a digital façade to cover existing structural malfunctions. In a country where the lines between power and accountability are already blurred, a tool like ChatGPT could be easily weaponized not to fight corruption but rather to cover it up.

This becomes much more relevant when considering the margin of error that AI technologies have when translating documents - especially those written in less commonly spoken languages like Albanian. AI systems have proven to be quite skilled at processing contemporary, widely spoken European languages like English, French, Spanish, and German. The efficacy of AI technologies is increased by their access to and training on large datasets. The same is not true for languages that are less common on the internet, like Albanian, because there is a significant lack of training data.

Although there are reasons to believe that ChatGPT will be providing the Albanian government with the best support, recent reports claim that ChatGPT 4.0 has been getting lazier. Even the official reply of OpenAI admitted that the behaviors of the AI model can be unpredictable. While humans are building increasingly complex AI models, we are becoming increasingly ignorant. How could we better ensure the quality and utilize the outputs of these generative AI models? The inexplicability may lead to two undesirable scenarios. First, humans take the outcomes for granted and stop questioning. Second, humans can manipulate the outcomes as, in any case, few, if not no one, can satisfactorily explain them. It requires both awareness and human effort to stay alert and discern. Therefore, we must resist the temptation to treat AI as a magic wand. For now, generative AI can be extremely useful in saving human efforts from administrative, mechanical, and descriptive tasks. The age for AI to help us accurately judge, interpret, and verify has not yet arrived.

Albania deserves the credit for thinking outside of the box, and as Dror Gill, a Generative AI consultant, put it, if Albania’s case is successful, it could become a model to follow for practitioners and authorities all over the world. However, if it fails, it could set back the trust in both AI and the EU. Because, at the end of the day, accession is about credibility, and it must be matched by real political reform, human oversight, and the recognition that no algorithm can replace diplomacy, dialogue, and democratic accountability.

 


About the Authors

Amy KOKALARI 

Amy Kokalari is a researcher and policy analyst specializing in European integration, digital governance, and the Western Balkans. She is currently pursuing a Master's in European Transformation and Integration at the College of Europe in Tirana, where she is researching cybersecurity resilience in the Western Balkans. She previously earned a Master's in Transnational Governance with a specialization in digitalization from the European University Institute in Florence. Amy has contributed to research at leading institutions, including the Centre for European Policy Studies (CEPS) in Brussels and the European University Institute. 

Her work has focused on EU enlargement policies, digital transformation, and geopolitical strategies. Beyond her academic and research engagements, Amy is fluent in multiple languages and passionate about tech policy and southeastern European studies.

Get in touch

 

 

Timothy YEUNG

Timothy Yu-Cheong Yeung is a Research Fellow at CEPS. He is also an Associate Researcher at the Leuven Center for Empirical Jurisprudence of KU Leuven in Belgium. He holds a Ph.D. in Economics from Toulouse School of Economics and a Master’s Degree in Economics from The University of Hong Kong. Timothy completed many research projects and published several academic papers in peer-reviewed journals, such as the Journal of Comparative Economics and Constitutional Political Economy. His research interests cover law and economics, digital economics, and EU institutions. 

He was invited to act as a member of the scientific committee of the book “Data Science for Economics and Finance: Methodologies and Applications” by the Joint Research Centre of the European Commission. He was involved in the setting up Governance Analytics/Chair of Governance and Regulation of Université Paris-Dauphine, PSL Research University. 

Get in touch

 

General News