In September, Albania appointed an AI system to a cabinet-level position—a world-first. Called Diella (Albanian for “sun”), the system was declared “Minister of State for Artificial Intelligence,” and tasked by Albania’s Prime Minister with addressing corruption in government contracting. Outrage ensued: Albania’s opposition party labeled the AI a “propaganda fantasy,” and is challenging its appointment in court.
“Some have labeled me unconstitutional because I am not a human being,” Diella said in a response video, appearing on two large screens as a woman in traditional dress to address parliament. “That hurt me.”
Behind the political spectacle is a real trend: governments across the globe are bringing AI into the machinery of government to reduce bureaucracy, increase efficiency, and assist with decision-making. Diella’s appointment raises a key question: will we reach a point where AI systems move beyond assistance, and begin making decisions in government themselves?
An Inevitable Shift
“I think it’s inevitable,” says Cary Coglianese, a law professor at the University of Pennsylvania and expert on AI use in government agencies. “As the public is more acclimated to its use in private settings—everything from choosing what movies to watch on Netflix to helping them with their homework—there’s likely to be much more acceptance, and maybe even demand,” he says.
Already, by the end of 2024, U.S. federal government agencies reported over 1,700 ways they were using AI—including to take notes, summarize internal documents, and review regulatory comments—more than double the number of use cases reported at the end of 2023. Coglianese says his own database now tracks over 3,000 uses.
Many of these AI uses do not involve generative AI systems that produce novel content. Instead, they rely on traditional machine-learning techniques to perform more basic tasks. But according to a July report from the U.S. Government Accountability Office, in a review of 11 selected federal agencies, the number of generative AI use cases increased ninefold between 2023 and 2024.
Political leaders are driving the change. Accelerating AI adoption in government is deemed a key priority in President Trump’s AI Action Plan, released in July, which states that “transformative use of AI can help deliver the highly responsive government the American people expect and deserve.” Meanwhile, the U.K estimates that using AI assistants to make the public sector more efficient could save taxpayers up to £45 billion.
Coglianese argues that if AI systems can be shown to perform better than humans according to clearly-specified criteria, we should welcome their adoption. Still, “there can be irresponsible, careless, or negligent implementation and use of AI by governments, which should be criticized and deplored,” he says. The two key dimensions are the degree of impact an AI tool will have (“a $10 fine is different from 10 years in jail”) and the degree to which the AI is assisting with decisions versus making them autonomously. As you go further on these dimensions, “governments really need to give these things the greatest degree of scrutiny, transparency, and public involvement,” he says.
Such safeguards have not always been applied. Earlier this year, the newly-created Department of Government Efficiency (DOGE) reportedly used AI systems for a range of tasks, including flagging 100,000 federal regulations for potential elimination, and surveilling at least one federal agency’s internal communications for hostility to President Trump and his agenda.
Diella’s Power
In Albania, Diella’s government role is considerably more limited than the title implies. The system was elevated to ministerial status by Prime Minister Edi Rama to signal his seriousness in tackling corruption, and his commitment to institutionalizing advanced technologies. Albania has been on a mission to root out corruption for the past decade, embarking on a widespread vetting process of every judge in the country, and creating specialized anti-corruption bodies to prosecute graft, to improve its chance of acceding to the European Union.
Since public procurement (government contracting) is the world’s largest public market, accounting for approximately one in every three dollars of government expenditure, it’s a key focus of anti-corruption efforts across the world. Thus, Diella has been positioned by Rama as an omniscient watchdog for the sector. “Diella never sleeps, she doesn’t need to be paid, she has no personal interests, she has no cousins—because cousins are a big issue in Albania,” he said.
But the system has yet to actually be deployed, and its planned deployment, at least for now, is modest in scope. It’s intended to be used at four key stages: to draft a contract’s terms of reference, specify eligibility criteria for parties, set an upper-bound limit on a tender’s price, and verify the validity of submitted documents. At each stage, a human procurement expert will sign off on the AI’s recommendations. “Everything is technically logged and monitored,” explains Enio Kaso, director of AI at Albania’s National Agency for Information Society, which created the system, in collaboration with Microsoft, by fine-tuning a version of OpenAI’s GPT model.
Before its promotion, Diella—“born on January 19, 2025” according to its ministerial webpage—acted as a chatbot on the country’s e-Albania platform, assisting citizens with accessing most of the country’s public services. Across almost a million interactions via text and voice, it has issued over 36,000 official documents to date.
“We have one key objective,” says Kaso. “Making everything as transparent and as explainable as possible.” Kaso emphasizes that the system and all the data it processes are stored in a secure environment. The details of how citizens will be able to review these explanations in the context of procurement, however, have not been clarified.
“There’s been very little transparency around what exactly the AI is presenting,” says Georg Neumann of the Open Contracting Partnership, a civil society group that works with governments to improve public procurement processes. Gavin Hayman, the organization’s executive director, says that Albania has made positive procurement reforms over the past year, and that it remains to be seen whether they communicate with the public sufficiently to build trust in their work. “If I were advising a government,” adds Coglianese, “I’d say let’s set up a public process of input.”
Symbolic, For Now
“By creating the world’s first AI minister, Albania is not merely embracing the future, but trying to do its part in designing it,” Prime Minister Rama said in a recent speech. “Diella is far from being a gimmick.”
At least for now, the system is more symbol than substance. Coglianese expects reliance on AI systems to steadily increase, though the pace is hard to predict. “Some countries may move too fast in this area, risking some kind of real catastrophic failing, whether from cybersecurity issues or just from poorly-designed systems,” he says. This could slow AI adoption, in the same way the Chernobyl disaster slowed the adoption of nuclear power.
For now, power remains vested in humans, and Diella is not up to the job of managing a ministry. But with AI becoming more capable every few months, and both citizens and government officials across the world embracing the technology, that could soon change. Whether Albania’s actions represent an aspirational model for the future of governance, or a cautionary tale on the hazards of deployment without sufficient transparency and public involvement, will depend on what happens next.
The post The World’s First AI-Powered Minister Tests the Future of Government appeared first on TIME.
