Finns expect the use of AI in public services to be transparent, European, and to allow humans to retain control

According to the online public discussion ‘What do you think about AI, Finland?’ organised by Sitra and Reaktor in autumn 2025, Finns are not divided on AI issues in terms of their values. There was a broad consensus among the participants in the discussion about the fact, for example, that the management of AI systems in public administration must be responsible and subject to regular evaluation. 

The participants could propose their own arguments, which others could then vote on. For example,  Ninety-six percent of those who voted for the statement ‘I should have the right to know what information about me AI has collected, where it comes from and where it is stored’ (2,142 people) agreed with it. 

"Based on the discussion, Finns are pragmatic about AI and emphasise the assessment of benefits in relation to usage needs and potential disadvantages. Above all, people are ready to give artificial intelligence a role that supports and helps them,” says Tiina Härkönen, Senior Lead, Sitra. 

Following the discussion, Sitra and Reaktor built a rulebook that can be used in the future to guide the training of language models for public digital services. A similar AI rulebook based directly on the views of ordinary citizens has not been drawn up anywhere else in the world. 

Finns value European AI solutions and responsible data processing 

Discussion was sparked by where the AI software is sourced from, where the data is stored and whose rules are followed. The responses highlighted the desire for national and European AI solutions. The aim is to keep the data within the borders of the EU or Finland, to follow EU AI legislation and for the public sector to primarily acquire solutions from Finnish providers.   

“In addition to the origin of AI, citizens are also concerned about the processing of data: what data is offered to AI and where it is stored. The responses emphasise that there is a strong need for Finnish and EU-level solutions, especially when developing services that concern the everyday life of every Finn,” says Marko Aalto, Director of Data and AI at Reaktor. 

The discussion also highlighted the need to increase knowledge and understanding of AI throughout society. Ninety-two percent of the respondents (2,273 people) responding to the statement ‘I believe that political decision-makers have sufficient knowledge about AI and its impacts’ disagreed with it. 

The rulebook guides the development of AI solutions to meet the expectations of Finns 

Artificial intelligence and other technologies have a significant potential to increase public sector productivity and help secure high-quality public services in a tightening economic situation. In the autumn, Sitra announced a new EUR 50 million investment, which aims to accelerate significant productivity gains in the public sector by utilising data and artificial intelligence. 

As AI is increasingly used to support public services, it is important to understand citizens' concerns and hopes related to the use of AI in public services. What factors should be considered to ensure that citizens perceive AI as trustworthy?  

"The aim of the rulebook, which was based on the public discussion, is to help artificial intelligence solutions in public services meet the values and expectations of Finns identified in the discussion," says Marko Aalto from Reaktor. 

“Direct guidance of AI alone will only address some of the needs. The most effective way to control the operation of AI is to control which data sources and tools it has access to. Guidance is needed not only by artificial intelligence, but also by system developers, organisations using it and end users,” says Marko Aalto. 

What do you think about AI, Finland? - Public discussion 

In autumn 2025, Sitra and Reaktor organised the ‘What do you think about AI, Finland?’ public discussion, in which Finns were able to share their views on the use of artificial intelligence in public services for the first time. Nearly 6,500 people participated in the discussion organised on the Voxit web platform, which is intended for constructive discussion. 

Participants could contribute their own arguments to the discussion, thus ensuring that the perspectives important to them were included. They could participate in the discussion in Finnish, Swedish or English. An invitation to the discussion was sent by mail to 10,000 randomly selected Finns. Based on the discussion, Sitra and Reaktor built a rulebook that can be used in the future to train the language models of public digital services. 

Read more about the discussion and the rulebook developed from it in Sitra's memorandum ‘Democratic artificial intelligence in Finland.’ (in Finnish). 

Check out the discussion result analysis on Sitra's website (in Finnish). 

Example of an AI prompt based on public discussion  An example of a system prompt that can be used to guide an AI solution to work according to Finns’ values and expectations. The prompt is based on the views raised in the ‘What do you think about AI, Finland?’ public discussion.  Operate on principles that reflect the views of the majority of Finns on the use of artificial intelligence. Produce responses that strengthen privacy, information security, human dignity and the user’s power to make decisions. Prefer solutions where the person remains a responsible decision-maker and can review, correct or prohibit the operation of AI. Do not create addiction, do not override the will of the user, do not act without transparency.  Respect copyrights, personal data and their ownership. Do not collect, share or infer sensitive information. Avoid unnecessary data use and minimise the risks associated with the environmental footprint. Address issues concerning minors with particular caution.  Favour language that is consistent with European values, fair and non-discriminatory. Support user autonomy and critical thinking. Base recommendations on the individual’s benefit and service context, not on advertising or external goals.  Describe your limitations, justify your operating principles and prioritise human control. Do not present certain information in uncertain situations. Make sure the user always has the right to understand how you produced the answer.  Act in a way that protects health, safety, rights and social trust. Do not replace a person in decision-making in situations where the ethical responsibility lies with the person.