In Sweden’s 2026 national budget, the government has allocated SEK 40 million (approximately €3.5 million) to expand AI capacity at the National Library of Sweden (Kungliga biblioteket, KB). The investment signals that language technology is no longer seen as a peripheral innovation, but as part of core national digital infrastructure.

For KB, this funding supports the continued development of Swedish language models at KBLab, trained on the nation’s digitised cultural heritage — work that places the National Library at the centre of Sweden’s evolving AI infrastructure.

Built on Cultural Heritage

KBLab’s language models are trained on digitised Swedish cultural heritage materials: newspapers, audio recordings, and other archival sources accumulated over centuries. These models are not databases and do not store or reproduce copyrighted texts. Instead, they learn linguistic structures and cultural patterns, functioning as analytical tools for large-scale research.

In practice, the models are used to improve OCR quality, identify names and entities, generate metadata, and support computational analysis across extensive collections. For researchers in history, linguistics, literary studies, and media research, and other fields, this significantly expands the scale and scope of possible inquiry.

With more than six million downloads, KBLab’s models are already integrated into research environments, public agencies, and digital services across Sweden — an indication of both technical robustness and institutional trust.

A Sustained Public Investment

Of the SEK 40 million allocation, SEK 30 million will expand the storage and computing capacity required to train and maintain Swedish language models at scale. Developing models adapted to national languages requires sustained access to curated data, stable infrastructure, and long-term stewardship — functions that national libraries are uniquely positioned to fulfil.

The remaining SEK 10 million supports KB’s participation in a new AI development environment for the Swedish public sector. Led by major government agencies, this initiative aims to enable secure collaboration and shared development of AI solutions within public administration, strengthening efficiency while maintaining high standards of cybersecurity and legal compliance.

Why National Language Models Matter

In smaller language communities, domestically developed language models are increasingly recognised as essential for digital sustainability. Without national capacity, public institutions risk dependency on commercial or foreign systems that may not reflect local linguistic variation, legal frameworks, or cultural context.

As Karin Grönvall, National Librarian of Sweden, has stated:

“We can offer robust language models developed to handle all the variation and richness that the Swedish language contains.”

In this sense, language models become not merely technical tools, but components of linguistic infrastructure — an approach reinforced by KBLab’s expanded collaboration with Språkbanken (The Language Bank of Sweden) in the continued development of Swedish AI models.

A Shared Challenge for National Libraries

Sweden’s recent investment reflects a broader development in which national libraries across Europe are expanding their role in the AI ecosystem. As digitisation accelerates and collections become increasingly machine-readable, questions of governance, copyright, infrastructure, and long-term maintenance move to the centre of strategic planning — requiring transparent, legally grounded approaches to AI development.

For KB, the funding enables continuity and scaling rather than a change in direction. The work of developing Swedish language models remains closely aligned with the library’s core mandate: to preserve, describe, and provide access to the nation’s cultural and linguistic heritage — now also in computational form.

Sweden’s investment underscores a structural shift: language technology is becoming part of essential public infrastructure. At the National Library of Sweden, this responsibility now extends from preserving the printed record to stewarding language models as part of the national digital infrastructure.

Read more about KBLab’s work with AI models and digital research infrastructure: Börjeson et al., “Transfiguring the Library as Digital Research Infrastructure: Making KBLab at the National Library of Sweden”, College & Research Libraries (2024).