It is against this backdrop that, since 2023, the Albanian government has embraced AI, launching initiatives to expedite the translation and consolidation of legislation, to review procurement opportunities, and appointing an AI system, Diella, to a cabinet-level position. While technological innovation appears to offer a pragmatic mechanism to stymie corruption, facilitate citizen participation, and document legislation, such tools risk addressing only surface-level symptoms instead of deeper ethical and institutional reforms necessary for compliance with the EU acquis.

The benefits of AI in the EU accession process

And such was the thinking when Rama introduced Diella, minister for public procurement and the world’s first AI minister. This cabinet-level position was assigned to an AI agent dedicated to evaluating procurement at multiple levels across the nation, deciding the recipients of public tenders, digitizing records, and translating laws. And some have suggested that the endeavour might yield encouraging results, particularly noting that it could help standardize a process that is frequently assailed by major corruption scandals.

The question, then, is not whether AI can help Albania. It is whether the government’s current version of AI integration, implemented rapidly, with unclear safeguards, and without significant involvement from opposition actors, independent institutions, or civil society, can contribute to the kind of rule-of-law transformation the EU requires. The problem is less technological than political: how can AI be deployed to strengthen democratic institutions rather than amplify their weaknesses?

Challenges for Albania and AI implementation

One of the first hurdles is that AI is not being introduced into a neutral landscape; instead, it is arriving in a context defined by institutional fragility, persistent allegations of corruption, and political mistrust. These conditions shape both the deployment and perception of AI. Because machine-learning systems often reproduce societal biases and reflect those biases back to users, they tend to create a cyclical environment that reinforces dominant social and power structures instead of challenging the inequalities embedded in training data. In the context of EU accession, this presents a distinct governance problem as biased or unweighted data may generate outcomes that conflict with EU legal obligations, including Article 21 of the Charter on Fundamental Rights that prohibits discrimination. Thus, in Albania, where judicial reforms remain incomplete, and the division between ruling and opposition parties is sharply polarized, algorithmic systems risk entrenching the status quo rather than challenging it. A system trained on uneven or politically skewed datasets may end up reflecting the very problems the EU accession process is meant to correct.

Additionally, while Rama’s ruling Socialist Party has extoled the positive impact AI might have in the government, the opposition Democratic Party is more wary, framing AI as a tool of the ruling party. This skepticism is in part due to the Socialist Party’s uninterrupted hold on power since 2013, along with its continued centralization of authority over weakened institutions, and cooptation of independent media. Additionally, recent elections have garnered concerns about government practices and the handling of the electoral process.          

Moreover, the implementation and supervision of Albania’s AI system falls under the direct supervision of the Prime Minister’s office and executive agencies like the National Agency for Information Society. There is no parliamentary committee or body responsible for the system, or a true mechanism for oversight and accountability. In this context, the introduction of AI risks further consolidating existing power asymmetries, transforming technology into a mechanism that reflects, and potentially reinforces, Rama’s grip over state institutions rather than catalyzing institutional reform.

AI and EU governance

A central feature of the EU’s digital governance model is its emphasis on process-based compliance. EU institutions assess not only whether technologies improve procedural outcomes, but also whether the systems are made transparent, auditable, and subject to institutional checks. The EU AI Act formalizes this logic by classifying many governmental uses of AI as “high-risk,” triggering stringent requirements related to data governance, explainability, accountability, and supervisory oversight. These requirements are intended to ensure that AI-enhanced decision-making remains compatible with rule-of-law standards rather than simply insulating political actors from scrutiny.

For EU candidate countries such as Albania, this regulatory model presents both an opportunity and a constraint. While digital modernization is often encouraged as part of administrative reform, the adoption of AI systems without parallel development of oversight bodies, enforcement mechanisms, and regulatory clarityrisks creating misalignment with EU standards. Policy analyses of AI governance in accession contexts emphasize that partial or symbolic alignment, particularly where AI systems are introduced more quickly than institutional safeguards, can complicate the accession process rather than accelerate it.

While Albania has committed in principle to aligning with the EU acquis as part of the accession process, its integration of EU frameworks remains partial and uneven. One recent assessment notes that Albania has yet to fully transpose or integrate core elements of the EU’s emerging digital rulebook, including comprehensive oversight mechanisms for AI systems. This gap raises concerns that AI adoption in government may outpace the institutional capacity needed to ensure compliance with EU standards, especially in high-risk domains central to democratic governance and public trust.

Moreover, in a political climate marked by distrust between ruling and opposition parties, AI implementation risks becoming another flashpoint. If the government controls access to AI-based legislative tools or deploys systems without cross-party consultation, critics may frame the technology as a partisan instrument rather than a neutral administrative aid. That perception alone could erode trust in the reform process, raising concerns in Brussels about the sustainability of AI-based institutional modernization.

Implications and implementation

This oversight body should mandate algorithmic impact assessments for high-risk public-sector AI systems (like Diella), require human oversight in decision-making functions, and establish a public registry of all AI systems used by state institutions. Such a registry should include information on system purpose/usage, oversight mechanisms, and audit outcomes. This would ensure transparency and enable civil society and citizens to better monitor how the government is using AI.

Moreover, AI governance should formally incorporate domestic stakeholders, including Albanian universities, civil society organizations, and journalists, into the design, evaluation, and monitoring of these AI systems. This would reduce over-reliance on executive, government agencies and external vendors while strengthening democratic legitimacy, Albanian domestic capability, and public trust.

Finally, AI governance requirements should be codified in primary legislation, clearly defining institutional responsibilities and equipping Albanian regulators with the authority and capacity to oversee the deployment of AI systems throughout government frameworks. Such legislation should also place meaningful limits on executive discretion, including the role of the prime minister, in the unilateral control or deployment of AI systems within public administration. Legislation should also establish clear provisions that identify who bears legal and administrative responsibility for the actions and outcomes of AI systems. Lastly, these laws should mandate a public articulation of the norms and values embedded in these systems, including clear explanations of how those values are chosen and weighted, whether that content comes from the Albanian government or partners such as Microsoft, and how current deployments, such as Diella, are aligned with Albanian law and societal expectations.

Perhaps most importantly, Tirana must acknowledge that AI is a complement to, rather than a substitute for, longstanding priorities such as anti-corruption efforts, judicial reform, and the depoliticization of public institutions.

Outlook

While Albania’s approach may seem novel, it is not the only country seeking EU membership to implement AI in government contexts. Other EU aspirants, like Ukraine and Moldova, are also using AI tools in government systems. This new frontier of governance presents both potential risks and substantial benefits depending on how such tools are deployed. Albania can act as an example, demonstrating how digitization may strengthen a country’s legal system, democratic standards, and government efficiency, while also revealing the potential pitfalls when public buy-in and broader political stakeholder participation are absent. As government AI adoption increases, the EU should work with candidate states, not only to enact AI legislation, but to help consider how AI might be operationalized to strengthen democracy, uphold the rights of citizens, and reinforce public trust and government accountability.

Albania’s embrace of AI is ambitious, imaginative, and in many ways admirable. It reflects a government eager to resolve issues that have repeatedly slowed its accession and harmed its citizenry. But technological acceleration cannot replace the slow, painstaking work of institutional reform. Whether this moment becomes an opportunity to modernize or another example of innovation outpacing oversight will depend less on the algorithms that are implemented and more on political will, regulatory capacity, and the willingness to treat digital transformation as part of the broader democratic reforms that EU membership ultimately requires.