FLASH
Shkruar nga Clara Riedenstein*
Artificial intelligence agents are replacing humans as public servants. This raises serious legal and practical questions.
Albania, the small Balkan country, announced a bold artificial intelligence initiative: it has built an AI-powered chatbot called Diella, who will serve as the country's Minister of Public Procurement.
From now on, the artificial agent will manage all bids for government contracts and purchases, from road works to public health services. Diella will make Albania “a country where public tenders are 100% free of corruption,” says Prime Minister Edi Rama.
Artificial intelligence agents are replacing human experts in politics, law and medicine. The Estonian government is testing the use of AI judges in civil disputes. It’s not just governments, users are following suit. A recent report found that more than 10 million Britons are using chatbots to manage and self-diagnose mental health problems.
These efforts seem reasonable, even a good solution to urgent issues. Humans are flawed. They are corruptible, slow, and often make mistakes. So it makes sense that they could be replaced with seemingly infallible beings.
However, while agents that replace human experts may initially seem appealing, reliance on artificial intelligence raises serious and as yet unanswered questions about accountability and effectiveness.
Motivations for turning to artificial intelligence chatbots range from increasing transparency to enhancing the delivery of public services.
Albania suffers from widespread corruption. The most recent report by the US State Department on the country concluded that corruption was present in all branches of government, “including through public procurement and public-private partnerships.” This is hindering the country’s accession to the European Union, a goal that Prime Minister Rama aims to achieve by 2030.
Other countries are exploring artificial intelligence solutions to their bureaucratic backlogs. In Estonia, the number of judges has remained constant over the past two decades, while the number of cases has almost doubled. In the United Kingdom, the Crown Court had a record 67,000 cases backlog at the end of 2024.
There is a growing need for mental health services. As of mid-2025, over a million people are on the UK's National Health Service waiting list to receive support.
Although appealing, this transition from humans to artificial intelligence experts comes with serious risks.
While chatbots are impressively reliable and autonomous, they are fallible. The problem is that when artificial intelligence makes a mistake, it's not clear who is responsible.
Large language models like ChatGPT or Diella have a neural network system modeled after the human brain. They are fed large amounts of data, which in turn allows them to make calculations between data points.
While this makes them good at making autonomous predictions and providing reliable answers, it is natural for such systems (just like our brains) that they sometimes give the wrong answer.
When human experts make mistakes, they can be held liable. For doctors and lawyers, professional misconduct law steps in to protect citizens when human experts act negligently or cause preventable harm. When ministers and judges make mistakes, public backlash can put pressure on governments to force them to resign.
Such accountability does not exist with artificial intelligence. If an AI agent makes a mistake, it is difficult to hold it accountable.
The legal issues are mounting. An Air Canada customer recently sued the company after its chatbot gave him false information about the carrier's airline policy. In response, the company argued that it could not be held liable for the advice given by the chatbot because it was a "separate legal entity that is responsible for its own choices."
Although the airline lost the case and had to compensate the customer, the bigger question of how to manage misleading or false responses from chatbots remains unanswered. Artificial intelligence chatbots are often called ethical “black boxes.” That’s because, unlike humans, they can’t explain their decision-making.
If the Minister of Artificial Intelligence, Diella, makes a public procurement decision that turns out to cause great harm, she will not be able to explain the reasoning behind why she made this decision.
This is an urgent question, as artificial intelligence agents replace the experts from whom citizens look for well-thought-out judgments and explanations.
Government officials empowered by artificial intelligence also pose practical problems. As Prime Minister Rama himself said in his announcement, “Diella is the first cabinet member who is not physically present.” She cannot sit at the table during cabinet meetings and cannot be shut down if she breaks the law.
While there are risks in replacing bureaucrats with AI agents, the technology could increase their productivity. In Germany, AI is being tested to assist lawyers and judges. This allows judges to focus on the complex questions in their work, where human judgment is required, and leave administrative tasks to AI.
The UK's National Health Service is experimenting with using chatbots to triage people seeking mental health support, so that those in need can find a human specialist as quickly as possible.
The appointment of an artificial intelligence minister from Albania made a splash in the media. The risks of increased accountability from artificial intelligence systems are real. But so are the opportunities for this technology to improve public services. /Adapted from Pamphlet/
*Clara Riedenstein is a technology policy analyst and writer whose work examines how emerging technologies shape existing political, legal, and social institutions. Her research has been published in Bandwidth, Tech Policy Press, and European View...
Lini një Përgjigje