IVA's President: Some thoughts on AI, Democracy, and Trust

Dear Friends of IVA,

Recently I took part in a discussion about AI and democracy on Swedish Radio. One thing struck me:

The debate on AI and democracy often revolves around information and power. We discuss what information is disseminated, how public opinion is influenced, and how economic and political power can shift between countries and major tech companies. Considerably less attention is paid to how AI affects the state itself, and how the state’s use of AI, in turn, affects democracy.

Mature democracies are built on the ability to gather information, process it, weigh different interests against one another, and translate this into decisions. This also happens to be a fairly accurate description of what AI is getting better and better at doing. AI’s greatest strength lies in its ability to process enormous amounts of information and draw conclusions. Used correctly, it can contribute to better and more cost-effective public services, strengthen societal resilience against internal and external threats, improve decision-making, and make it easier for people to exercise their rights and fulfill their obligations as citizens. In other words, AI has great potential to strengthen democracy. But when the public sector fails to keep pace with developments, AI can instead become a threat to democracy. 

Sweden’s AI Commission noted in its report to the government that it is hard to regulate a technology that is not yet fully understood. This is partly due to what is known as the “Collingridge Dilemma”– the fact that the government has the greatest power over a technology when it is in its infancy, but that is also when the government understands it the least. However, this is also because the pace of AI development – and its far-reaching impact on virtually every policy area – exceed the government’s capacity to manage it.

The need for the government to swiftly build expertise and practical knowledge is also one reason why the AI Commission, of which I was a member, delivered its recommendations six months ahead of schedule: to underscore the urgency for the government to embrace the technology in a timely and appropriate manner.

Sweden has built its success on strong institutions, high levels of trust, and a social contract in which people accept change because they trust that society can handle transition. Historically, Sweden has been good at managing structural change. We have invested in people’s ability to adapt and have sought to ensure that the benefits of progress are shared by many. This is an important part of the Swedish model. But perhaps the most important strength of the Swedish model is that it has fostered trust – in technology, in the government and its ability to manage transition, and in the future.

In several countries, we are now seeing growing concern about AI alongside declining confidence in the ability of the state to manage it effectively. This is particularly true of the United States. In an international survey of 21 countries in 2025, the US had the lowest share of positive responses regarding attitudes toward AI. Two-thirds of Americans said they were concerned about AI. At the same time, 31 percent responded that they had no confidence at all in the government’s ability or willingness to manage AI development for the good of society.

In another survey which included Sweden, the country ranked at the lower end of those polled when it came to public confidence in the government’s regulation of AI.

With its tremendous pace of development and scope, AI fundamentally challenges the Swedish model – and perhaps especially the public sector’s ability to manage the technology for the benefit of society. In doing so, it puts trust to the test. I have previously written about “Axel Oxenstierna 2.0” and the need to adapt the governance model, which has served Sweden so well, to an era of complex and interrelated problems and an accelerated pace of change. For this reason, IVA has established a working group of experts that will issue recommendations around the turn of the year.

To ensure that AI strengthens democracy, Sweden must therefore do two things at the same time. We must be fast enough to deploy AI where the technology can create value. And we must be wise enough to ensure that its development proceeds in a way that strengthens trust rather than erodes it. In both cases, public administration plays a crucial role. The greatest risk is not that we move too quickly. The greatest risk is that we are passive, leave development to others, and lose the ability to influence it.