The Young Academy of Sweden’s response to the AI Commission’s “Roadmap for Sweden”

The Young Academy of Sweden has submitted its consultation response to the report SOU 2025:12 AI Commission’s Roadmap for Sweden. We share the Commission’s overall ambition, but have critical objections and recommendations.

In Swedish: Response regarding SOU 2025:12 The AI Commission’s Roadmap for Sweden (PDF)

In December 2023, the government appointed the AI Commission, with a stated mission to provide concrete proposals on how artificial intelligence (AI) could promote Sweden as a leading research and development nation. After over a year of work, led by former Volvo Chairperson Carl-Henrik Svanberg, the commission delivered its roadmap. With the help of a series of proposed initiatives, Sweden would now assume a leading role in international AI development, with a specific focus on societal benefit, sustainability, and security. So far, so good – but the commission’s proposals risk stifling what has given AI its strength: free research and academic excellence.

AI is not like any other technology. It is transformative and affects all parts of society. Implemented optimally, AI also has the potential to solve some of our time’s major challenges: an effective climate transition, personalised healthcare, and reduced global poverty. To reach this potential, interdisciplinary, careful and analytical work is required, with close collaboration between research, industry, and business.

From a research perspective, we see primarily three problems in the commission’s proposals: excessive detailed control of research, a deficit of expertise at the decision-making and evaluation stage, and that access to data is a bottleneck.

We propose three clear solutions:

1. All research funding must be distributed through open calls. Let Sweden’s best researchers – regardless of institution or organisation – compete for funds based on scientific quality. Similarly, research must be protected from political micromanagement. Innovation does not arise on command – it emerges in freedom and with long-term investments.

2. Academic expertise must be represented in the proposed key functions. It is the researchers and academia that have developed the AI we are using today, and who can also lead development where AI can be beneficial. It is also the researchers who are at the forefront of AI development and have knowledge of its possibilities, and risks. Therefore, Sweden must seek assistance from the university sector to make the right decisions for the future.

3. Investments in data infrastructure must be made in close consultation with researchers to ensure both access to data and ethical data management. The development of socially beneficial AI is directly dependent on adequate data access – something that is currently partially restricted by existing legislation where certain data cannot leave certain predetermined physical locations. Similarly, the development of secure AI requires that data be made available in a way that maintains ethical provisions regarding patient safety and source protection. The issues that currently exist regarding research’s access to and handling of data must be reviewed – not least in dialogue with active researchers.