Albania's new AI bot minister: hype or help?

• Virtual minister sparks debate on rule by AI
• Bias, hacks and flawed algorithms among risks
• EU membership requires alignment with AI rules

BRUSSELS - Albania's new public procurement minister, an AI-generated bot named Diella, may be impervious to bribes and political pressure but experts doubt the world's first virtual cabinet member will be immune to mistakes or manipulation.

"This is huge and it could actually set a precedent," Bojana Zorić, a policy analyst at the European Union Institute for Security Studies told Context.

"Whether it's dangerous or not, we'll see."

Appointed by new Prime Minister Edi Rama in September, Diella, which means sun in Albanian, has been charged with managing all public government tenders to private companies.

Such tenders have previously led to some of the many corruption scandals that have stymied the Balkan nation's bid to join the EU. Rama has pledged to achieve EU membership by 2030.

Tech experts say outsourcing government decisions to an algorithm comes with risks and question whether Diella can really be relied on to eliminate corruption.

Here's what you need to know:

Can an AI bot stamp out corruption?

Diella was launched this year as an AI-powered virtual assistant on the e-Albania platform to help citizens and businesses with tasks such as obtaining state documents.

By September, it had registered more than 900,000 interactions and 36,000 documents carried its digital seal, according to a government website.

But promoting the bot to a decision-making role could be risky, tech experts say.

"It's a technical system like any other and is open to manipulation and corruption," said Joe Litobarski, a PhD candidate researching the history of electronic democracy at Maastricht University in the Netherlands.

While automated decision-making can boost efficiency, experts say that AI can produce unreliable results due to biased input data, flawed algorithms or irresponsible implementation.

AI needs transparency, human oversight and robust testing, according to global anti-corruption nonprofit Transparency International.

"It should not be seen as a substitute for governance, legal frameworks, (and) even human judgement," said Zorić.

What are the main concerns about an AI-generated minister?

Using AI to govern could heighten prejudice and increase data breaches, according to a 2024 report for the Organisation for Economic Co-operation and Development (OECD).

The public sector should test AI systems before implementation, train civil servants in digital skills and set up oversight mechanisms, the report said.

Experts also question who will be held accountable for any erroneous calls made by AI - pointing to autonomous vehicles that caused accidents and biased algorithms that falsely accused claimants of benefit fraud.

Tech experts also warn of the risks to Albania's digital sovereignty - how it controls its digital infrastructure and services - as foreign-made AI systems are more vulnerable to cyberattacks.

Tirana, notes that Diella was home-made, in collaboration with Microsoft, using its Azure OpenAI models.

Will Diella help Albania win EU membership?

The appointment of a "virtual entity" is noted in a European Commission's report on Albania's membership progress.

When questioned about the newly-appointed bot, a Commission spokesperson said the makeup of Albania's government was a "sovereign decision on which we have no comment".

Albania ranks among the most corrupt countries in Europe, according to Transparency International, and some experts say Diella indicates a genuine desire to tackle graft.

The Commission said Albania still needed to do more to fight corruption, organised crime and money laundering.

The Commission will continue to monitor progress, notably Albania's alignment with the EU AI Act, which classifies AI systems based on their potential for harm, the spokesperson said.

This means developers of AI systems like Diella must conduct risk assessments to root out bias and manipulation, monitor cybersecurity threats and keep records on how the system works.

Is this the future of government?

Countries from Canada to South Korea are already investing in AI to improve public sector services.

Britain's Prime Minister Keir Starmer has pledged to turn his country into an AI "superpower".

London is exploring the use of Anthropic's chatbot Claude for public services and has said that AI assistants had saved government coders the equivalent of 28 working days a year.

Some experts cite Estonia as an exemplar with tried-and-tested digital services linked to trusted digital identities.

Others say risks multiply if the drive to greater efficiency trumps all else.

"When it comes to modernising government you don't really want to run it like a start-up," said Litobarski. "I’m all for experimentation but let’s do it in a controlled, evidence-based way."

(Reporting by Joanna Gill in Brussels. Editing by Lyndsay Griffiths.)