In September last year, Prime Minister Edi Rama attracted the attention of international media with the appointment to his cabinet of "Diella" - a virtual chatbot integrated until then into the e-Albania platform of digital government services.

According to him, "Diella" would play the role of Minister of State for Artificial Intelligence with a specific task; freeing public tenders from corruption.

But while the sensation has faded a year later, her specific role still remains unclear.

At the Public Procurement Agency, the institution where Artificial Intelligence is expected to be tested, the work still appears to be in the preparatory phase.

“There is a project to develop an AI procurement system, but it is still in process,” said Reida Kashta, director of this agency, when asked if the agency had started using Artificial Intelligence in public procurement.

According to her, "the drafting of the terms of reference (ToR) has just been completed", the document that sets out the technical criteria for opening the tender that will contract the company responsible for building the system.

The public procurement system has been considered for years as one of the sectors most exposed to corruption in Albania. SPAK investigations have uncovered in several cases tender manipulation schemes, where the winners were predetermined before the procedures were carried out, while contracts for the construction of roads, incinerators or other public investments have become the subject of investigation for suspicions of favoring economic operators and violation of competition rules.

The government's ambition to integrate Artificial Intelligence into this sector is viewed with skepticism by the opposition and public policy experts due to the security risks it poses, the lack of human capacity, and legal uncertainties.

Democratic Party MP Klevis Balliu told BIRN that the use of AI can serve as an auxiliary tool for the administration, but cannot replace the legal responsibility of institutions or guarantee the fight against corruption on its own.

According to him, without a clear constitutional and legal basis, the establishment of a "virtual ministry" creates uncertainty over responsibility for decision-making, especially in a sensitive sector like public procurement.

"The sun does not solve corruption. On the contrary, it raises a fundamental question: who is responsible for decisions that affect a system that has no legally defined status?" Balliu said.

According to him, a virtual system can help increase efficiency and identify irregularities, but "it cannot replace transparency, control and human accountability."

Even public policy experts view with concern the use of Artificial Intelligence in procurement without a complete legal, institutional, and technical framework.

Alba Brojka, an independent expert on digital policies, says that Artificial Intelligence cannot be treated as a "magic wand solution" for complex problems like corruption, but only as a tool for increasing efficiency and automating administrative processes.

"It's like a knife, which can be used to cut bread or attack people. Especially since according to UNESCO's principle on the ethical use of AI, decision-making ultimately belongs to the expert/human and not to the technology," said Brojka, describing political speeches about the fight against corruption as unsupported.

In this context, she sees the use of AI as “a justification for a multi-year problem”, without first fixing the 'legal gap and institutional capacities'.

Even for Aranita Brahaj, director of the Albanian Institute of Sciences (AIS), the use of AI in procurement cannot be seen as a solution to problems with tenders without consolidating the legal basis and preparing a new methodology to monitor the application of these systems and their efficiency.

"Albania still does not have a complete sub-legal framework and, above all, this process comes at a time when public contracting is showing significant problems," said Brahaj.

Multiple challenges

In line with the ambition to introduce digital systems into public policies, the government amended the "Public Procurement" law in 2024 to include provisions that enable the use of AI in procurement processes.

Advanced technology systems are expected to be used primarily for drafting requests in the pre-tender phase, supporting the dynamic purchasing system and digital auctions, operating the Digital Complaints System, and monitoring the implementation of contracts.

The Director of the Public Procurement Agency, Kashta, says that this law is the legal basis for the use of AI in the procurement process.

But experts believe that these legal provisions do not create a complete basis that responds to all the challenges posed by the use of digital technologies, while a government strategy for the use of AI and a new law, which is still in the consultation phase, have yet to be adopted.

"These key horizontal acts, beyond sectoral legislation, should be in force and also, the institutions defined in the draft law on AI should have clear competencies and implementation capacities, so that we can effectively start implementing AI in public procurement," says Brojka.

According to her, this would guarantee control mechanisms, independent monitoring, licensing and registration of AI systems according to the categorizations that the law defines, auditing of algorithms, as well as effective mechanisms to denounce violations of rights.

But even the adoption of the new law does not solve the security issues associated with the implementation of AI systems, as according to Brojka, institutional capacities for data protection are lacking.

"We are faced with a lack of capacity from the Data Protection Commissioner to monitor private, public actors or different platforms, and it would be an even greater challenge to ask the same institution to monitor data for actors using AI, as this is not realistic," she added.

Brojka speaks of "multiple challenges" in applying AI systems in public policies and appeals for the application of AI to go hand in hand with the accountability and transparency of institutions.

She identifies as fundamental challenges the lack of an analysis that defines the concrete problems that the technology should solve, the quality of the data and the capacities of the PPA to understand, train and monitor the models. She also highlights the need for clear methodologies of responsible use and auditing, underlining that, despite the assistance of AI, the responsibility for the final decision-making in procurement must remain with humans.

"All of these are challenges that remain to be addressed by the PPA, and it seems that the institutions do not yet fully understand this preparatory cycle, but rather are driven by a political narrative, which is far from the conditions that need to be met for a responsible use of AI," Brojka emphasized.

Brahaj also warns that implementation will take time and requires monitoring. Equally important, she says, is how the evaluation, qualification and risk detection algorithm will be applied in tendering and how ready the business is to compete and appeal under AI application conditions.

Also, Brahaj adds, internal audit and the Albanian Supreme Audit Institution as an external auditor will need to develop a new methodology.

"Otherwise, this will remain a failed reform, where Albanian institutions will simply create a set of problems and defects that will serve as a 'laboratory' to teach others," Brahaj concluded. /BIRN/