Ursula von der Leyen’s tour of the Western Balkans this week came as the EU’s enlargement agenda hangs in the balance. With the enlargement package due next month and the Berlin Process summit and the European Council meeting days away, the trip captured the gap between Brussels’ stale rhetoric and the region’s expectations. Nevertheless, at the Tirana EU–Western Balkans Investment Forum, the European Commission president struck a note of recognition rather than ritual, hailing Albania as one of the EU’s most dependable partners, spotlighting AI as a key pillar for integrating the region into the Single Market — in tune with Albania’s own recent digital ambitions. Last month, Albania made global headlines as the first country to appoint an AI system as a cabinet minister. On 11 September 2025, Prime Minister Edi Rama unveiled Diella, a digital avatar of a woman in traditional Albanian costume, as State Minister for Artificial Intelligence, in charge of overseeing public procurement.
Zooming out, this move seems a bold attempt to showcase innovation and EU-minded modernity at a time when Tirana aims to open all remaining chapters in the EU accession process by year’s end. AI is also expected to facilitate Albania’s EU accession. But explicitly entrusting governance functions to a by definition ‘non-human’ algorithm-driven technology – possibly developed by non-European companies – raises critical issues of ethics and accountability, as well as concerns over cybersecurity, and technological dependency.
AI as the way to speed up EU accession?
Over the past few years, Albania has advanced steadily along its EU accession path, with 28 of 33 negotiation chapters already opened and only Cluster 5 on resources, agriculture and cohesion still pending. The country has set the goal to open this final cluster by the end of the year and to close all chapters by 2027. To reach this milestone, Prime Minister Rama intends to rely heavily on AI, which could advance Albania’s accession on three fronts. First, the development of secure digital infrastructure and the broader digitalisation of Albanian society are requirements for EU accession. The adoption of AI by Albanian public services feeds directly into this ambition, and on 1 January 2024, a ‘virtual public servant’ was launched on the e-Albania platform, essentially a chatbot assisting citizens with administrative procedures.
Second, AI has also become a tool to accelerate the accession process itself, by facilitating the transposition of the EU acquis and aligning Albania’s legislative framework with European standards. In 2023, the National Agency for Information Society (AKSHI) launched a €2.6 million tender for the ‘Use of Artificial Intelligence in the Process of Transposition of the Acquis for European Integration’. Prime Minister Rama personally reached out to then–OpenAI CEO and Albanian-born Mira Murati, striking a deal to use ChatGPT to translate and streamline the incorporation of thousands of pages of EU legislation into Albanian law.
Finally, AI has emerged as a potential anti-corruption tool. Considered by over one third of Albanian citizens as the greatest threat to the economy, corruption is not just a domestic concern but a key obstacle to EU accession. The 2024 Progress Report underlines that ‘corruption remains a serious concern [in Albania] and preventive efforts have had a limited impact’. Albania scored 42 out of 100 on Transparency International’s Corruption Perceptions Index 2024, ranking 80th of 180 countries, while one in four public service users admitted to paying a bribe in the previous 12 months. This is how Diella came into being, in an attempt to curb corruption through algorithmic rigor and supposed impartiality.
When the technological fix becomes the risk
The pervasive deployment of AI has nevertheless raised concerns. While internet take-up is relatively high in Albania, with 88% of people online, only 32% possess basic digital skills and 9% demonstrate above-basic skills. Albania should thus continue improving its performance on the EU Digital Economy and Society Index (DESI). But it must avoid rushing into AI-led governance without the institutional resilience, technological and infrastructural capacity, or social legitimacy required to manage the associated risks.
1 - Legitimacy risk
A key issue in introducing AI into governance lies in accountability and public trust. The government has repeatedly emphasised that Diella would not operate fully autonomously but under human supervision, with decision-making remaining formally in the hands of elected officials. Yet relying on information generated and shaped by AI is not neutral. Diella will inevitably sort, prioritise, and frame the information on which political decisions are based. The question is therefore to what extent human judgments are shaped, constrained, or biased by these outputs. This also raises the issue of responsibility: Diella will take countless ‘micro-decisions’ beyond direct political oversight, yet these will ultimately guide governmental choices. How traceable are such ‘micro-decisions’, and how transparent or auditable are the underlying AI models? Without clear answers, AI risks adding a further layer of opacity to governance rather than reducing it. This danger is amplified in Albania, where 75% of citizens distrust political parties and 60% distrust Parliament. The introduction of a virtual AI minister could further erode, rather than restore, public trust.
2 - Cybersecurity risk
Scepticism may also increase due to the cybersecurity risks inherent in deploying Diella. By appointing Diella as a fully-fledged minister, the Albanian government has increased its visibility and, with it, the likelihood of drawing the attention of potential attackers. Although its autonomy is limited and its purpose largely symbolic and image-driven, Diella is indeed a high-value target. It will handle sensitive personal and governmental data. This could encourage attackers to exploit the virtual assistant as an entry point into government systems via internal APIs, or to attempt direct access to data through social engineering, prompt injection, or compromised credentials.
3 - Risk of reliance on foreign AI
In the face of rising global tensions and Donald Trump’s weaponisation of Europe’s digital dependencies, the EU has become acutely aware of the strategic vulnerabilities created by its reliance on foreign technologies. Reducing these dependencies has now become a priority. Yet, although no official document confirms this, several indications suggest that Diella is based on OpenAI models and hosted on Microsoft Azure. This risks reinforcing the very dependencies the EU seeks to reduce, raising doubts about whether such reliance supports or complicates EU accession. The forthcoming Apply AI Strategy is indeed expected to underscore Europe’s ambition to reassert its technological autonomy vis-à-vis major American and Chinese providers. As Ursula von der Leyen noted during her visit to Tirana, Europe is already investing in this vision, building a network of AI Factories to help European start-ups develop, train, and deploy their own models. Crucially, this initiative will also extend to the Western Balkans, starting with two Factory Antennas in North Macedonia and Serbia.
How to keep the digital leap from becoming a digital trap?
Diella’s elevation builds on earlier AI initiatives such as her initial role as a virtual assistant within e-Albania, reflecting both a broader drive towards digital governance and a degree of political showmanship – yet still carrying risks of limited oversight and transparency. For the EU, Albania’s appointment of an AI Minister represents not only a governance dilemma but also a security concern in its immediate neighbourhood. As candidate countries advance along the EU path, the introduction of AI into public administration must be carefully managed: high-risk applications, including in public procurement, should always include human sign-off and robust human-in-the-loop oversight. Beyond procedural safeguards, cybersecurity should be strengthened to protect sensitive data from both internal errors and external threats. In parallel, supply-chain and cloud dependency risks associated with foreign AI models need to be addressed, as digitalisation should not come at the expense of autonomy. In line with DESI benchmarks, investment in digital skills and institutional capacity should also be enhanced, ensuring that citizens can trust AI-enabled services and that public institutions can respond effectively to emerging cyber threats. Without these safeguards, what is hailed as a digital breakthrough could turn into a cautionary tale of risky digital shortcuts.
