In September of this year, following a presidential decree that authorized Albania’s prime minister Edi Rama to create an AI-based virtual ministry, Rama appointed Diella as “Minister of Public Procurement.” With this decision, Diella became the first AI system in the world to hold a cabinet-level role. According to Rama, Diella’s mission is to oversee public procurement and fight corruption, aiming to bring transparency and efficiency to the public sector.

For Albanian citizens, Diella is not a new name. Since January 2025, she has been launched as a virtual assistant integrated into the government’s digital platform, e-Albania. Through her, citizens receive instructions on how to use online services and assistance in applying for or downloading digital documents. Although AI systems are used in many countries around the world for various state functions, Diella is the first case of an AI system holding a ministerial position.

Given the speed with which AI has become part of our daily lives, Diella may even seem like an ideal minister — one who operates on the basis of algorithms, does not take bribes, does not favor relatives, does not attend electoral dinners and does not demand rewards for political loyalty.

Diella is not affected by fatigue. Its analytical capacity is many times greater than that of a human team. An algorithmic system like Diella can analyze thousands of tenders in minutes, compare offers, detect price anomalies and identify hidden connections between companies. Processes that take months in traditional audits can be reduced to mere minutes.

As a result, the use of such an AI system is expected to reduce administrative costs, accelerate audit processes and strengthen transparency in the most vulnerable sector of public finances. In the best-case scenario, every actor in a tender is treated according to the same criteria, avoiding arbitrariness. Thus, Diella can ensure efficiency, impartiality and an unparalleled capacity to identify abuses. For a country like Albania, which ranks 80th out of 180 countries on Transparency International’s corruption index, this offers significant hope for curbing corruption in public procurement.

Despite the benefits that Diella could, in principle, bring, there are serious concerns about how this algorithmic system will function. In the worst case, the experiment risks eroding the foundations of democracy and violating Albania’s digital sovereignty.

Algorithmic corruption?

There is a widespread perception that algorithms are objective simply because they are technological. In reality, every AI system reflects the data it is trained on. If that data is biased, inaccurate, or incomplete, the system will reproduce those patterns. For example, if the Albanian public procurement data used to train Diella contains instances of abuse, corruption, discrimination or prejudice, the system risks mirroring and reinforcing them.

In other words, the conclusions and decisions of AI systems depend on the data they receive. As a result, the system can generate discriminatory or biased outcomes. To citizens, Diella’s decisions may appear “objective” on the surface, while in fact carrying invisible biases that risk leading to wrongful decisions.

Moreover, Diella’s source code — the set of instructions written in a programming language that defines the logic and operation of the system’s algorithms — is not transparent. Source code is the fundamental component that underpins the training, modification and development of an algorithmic system. When an AI system’s source code and operation are not transparent and cannot be reviewed by independent experts, there is no way to identify and correct potential errors or security vulnerabilities. This increases the risk of technical failures and makes the system less reliable. The lack of transparency around Diella’s source code renders its decisions technically inexplicable and makes the entire system suspect.

Thus, if there is no transparency regarding an AI system’s training data, algorithms and source code, the corruption present in Albania could be replaced by “algorithmic corruption,” which is much harder to detect and explain. We could move from a corrupt human system to one that is nontransparent and pseudo-objective. This presents a major technical problem, because not even Diella’s operators — much less ordinary citizens — can understand its decisions.

Another risk is the erosion of accountability. Democracy rests on a chain of responsibility: citizens elect parliament, parliament oversees the government and ministers are held accountable for their decisions. But what happens when an AI system like Diella makes those decisions? A ministry can be called to an interpellation, dismissed or prosecuted. Diella, by contrast, is not questioned, does not explain its decisions, and bears no political or legal responsibility.

At present, Albania lacks a legal framework that assigns liability to AI for wrongful decisions that may violate citizens’ rights. This creates uncertainty over who bears responsibility: if a tender is unfairly awarded, who can be sued? Who appears before parliamentary committees? A foreign team of programmers? Microsoft? Or perhaps an international NGO involved in Diella’s development? In such cases, responsibility becomes diffuse, turning AI into a tool for evading political accountability.

In the decree appointing Prime Minister Edi Rama, President Bajram Begaj stipulates in Article 2: “The Prime Minister also exercises responsibilities for the establishment and functioning of the Virtual Minister of AI ‘Diella’.” The same language appears in Article 1 of the decree on the composition of the Council of Ministers. In practice, this can be interpreted to mean that Rama himself bears political responsibility for Diella’s actions.

The appointment of an AI system as Minister of Public Procurement was strongly criticized by the Albanian opposition. Representatives of the Democratic Party (DP) argued that “Diella” would be used to evade responsibility for possible violations in public procurement.

A more democratic and legitimate approach would have been to establish a legal basis for her actions and decisions before the appointment. If, in the future, Diella makes arbitrary, incorrect, or harmful decisions for unknown or unpredictable reasons, the question arises: how will Prime Minister Rama explain and justify these actions before the Albanian Parliament and the public?

Without a clear, detailed legal framework defining Diella’s powers, responsibilities and limits, political responsibility falls entirely on Rama, exposing him to legal, political and public criticism. This situation not only jeopardizes institutional transparency and accountability but may also undermine citizens’ trust in public procurement decision-making processes.

Therefore, creating a law that clearly defines Diella’s powers, responsibilities and limits is essential. Such a step would prevent misunderstandings and strengthen its institutional legitimacy.

Violation of digital sovereignty

An additional concern is that Diella is developed on Microsoft’s large language models, an international company’s technology. This means decisions affecting millions, if not billions, of lek in public funds will be processed and filtered through algorithms developed with the help of a foreign corporation. This raises a fundamental question: Is it acceptable to delegate such an important and sovereign function — management of public money — to an international company?

The problem is not only the use of foreign technologies, which is common today, but the potential dependence of a state on a digital infrastructure it cannot fully control. Who can guarantee, for example, that the algorithmic logic on which Diella operates is completely impartial and protected from foreign commercial or political interests? What happens if a private company, even indirectly, manages to influence how public resources are distributed through the algorithm’s technical parameters?

If Albania relies to this extent on foreign platforms for critical functions, the country risks losing some control over its data and decision-making processes. In this case, we are dealing with extremely sensitive data: public tenders, the budget and state spending — elements that directly affect the functioning of the state and economic stability.

Moreover, the risk of setting a precedent is great. If today an AI tool like Diella is limited to public procurement, tomorrow a similar system could extend to even more important areas — such as national budget decisions, setting social policy priorities, or even initiating legal reforms. In such a scenario, a state’s digital sovereignty is seriously called into question, as political decisions begin to be influenced by algorithmic systems essentially designed, developed and controlled by private foreign companies.

In this way, the state risks becoming dependent on foreign companies’ technology. In the long run, this means that control over national data, the logic of fund distribution and future policies may no longer be managed exclusively by domestic institutions, but conditioned by external technologies and interests.

This discussion is closely tied to the concept of national security. Just as a state does not delegate its military or national defense to a third party, it should not allow the management of state finances and strategic data to be controlled by external actors. Otherwise, the country risks exposure to manipulation — economic, political or even geostrategic.

The issue becomes even more sensitive when we consider cybersecurity. Once Diella becomes the heart of public procurement administration, it automatically becomes a prime target for cyberattacks, whether by other states, criminal groups or local and international competitors. If Diella were attacked and paralyzed, all public procurement would be jeopardized: administrative processes would be blocked; sensitive data on contracts and the budget could be stolen, manipulated, or destroyed; and, above all, a deep crisis of confidence would emerge among citizens and businesses, who would see the state as incapable of protecting public finances and information.

It is enough to recall 2022, when Albania itself fell victim to a major cyberattack that paralyzed digital public services for weeks. This precedent shows clearly that the threat is not hypothetical but a tangible reality that could recur and future attacks could target far more sensitive systems, such as Diella. International experience only reinforces this concern.

Estonia faced a massive cyberattack in 2007 that paralyzed banks and state institutions, while the United States experienced an energy crisis in 2021 after the attack on Colonial Pipeline. These cases show that even the most powerful states are technologically vulnerable. If countries like Estonia and the United States, with advanced digital systems, have not been fully protected, then a state like Albania, with limited technological and financial capacities, is even more exposed. Such a situation constitutes not only a violation of digital sovereignty, but also a serious threat to economic and political stability and to national security as a whole.

Digital sovereignty should therefore be seen not merely as ownership of technology, but as a necessary element of protecting critical national infrastructure. Without a clear cybersecurity strategy and mechanisms that guarantee full national control over the functioning of such systems, any attempt to digitize public administration remains fragile and at risk. In this sense, Diella is not only a technological project to combat and prevent corruption, but also a major test of Albania’s ability to maintain its digital sovereignty.

Diella represents a bold experiment that puts Albania in the international spotlight for government-level use of AI. Its potential to increase transparency and combat corruption in public procurement is undeniable, but so are the risks posed by opaque algorithms, unclear political and legal accountability, and dependence on foreign technologies. For this reason, technology should serve only as an auxiliary tool to improve and accelerate institutional work, not as a substitute for political decision-making.

Ensuring accountability and legitimacy requires a clear legal framework, detailed reporting to the Assembly, independent auditing of algorithms and the safeguarding of digital sovereignty. Only in this way can Diella become a success story in the fight against corruption without undermining democratic foundations or exposing the country to foreign technological dependence and cybersecurity risks.

Feature image: K2.0

 Want to support our journalism? Become a member of HIVE or consider making a donation. Learn more here.