2 of 14

Capital: Oslo��Population: Approx�5.47 million (2023)��Currency: Norwegian �Krone (NOK)��Government: �Constitutional monarchy��National Day: May 17^th��Official Languages: Norwegian, Sami��Geography: Norway has the world's second-longest coastline, around 100,000 km, and is known for its deep fjords formed during the last ice age. The country has a varied climate, with a warmer coastal climate due to the Gulf Stream, a colder inland climate, and an arctic climate in the north�

NORWAY

3 of 14

Why is it natural for the National Library to work with AI?

4 of 14

The misson

The National Library's mission is to guarantee the delivery and preservation of published works across all publishing platforms, positioning itself as the primary source of knowledge regarding Norway and its circumstances.
The National Library promotes and provides access to cultural heritage, serving as a resource for research, education, and language development, while also helping to foster identity and a sense of belonging.
The National Library functions as both a support system for Norwegian research and is recognized as an independent research entity.
Through the Language Bank, the National Library provides datasets related to the Norwegian language for text and speech, aimed at advancing language technology in Norway.
Norway has a significant digitization history that remains largely unknown to many.

5 of 14

Some figures from the digital collection – 20 years later�

Books: around 630,000

Newspapers: about 4.5 million

Magazines: roughly 200,000

Letters and manuscripts: approximately 30,000

Sheet music: close to 6,000

Music manuscripts: nearly 10,000

Images: around 2.2 million

Posters: about 10,000

Maps: approx 3,000

Radio: roughly 2.2 million files

Music: about 285,000 files

Film: around 17,000 files

Television: approximately 3,800 files

�

2005: The collection must be digitized and made available on the Internet.

�

These are digital data about Norwegian language, culture, art, history, geography, and social life – collected in one place, which can be used to train computers to understand "Norway and Norwegian affairs."

6 of 14

Large Language Models

Challenges

Language models are increasingly essential in various fields.
They could greatly affect social interactions, highlighting the importance of assessing associated risks.
Major tech companies and large language communities lead in software creation, model training, and online services.
There is limited transparency regarding the training and configuration of models that power services like ChatGPT.
We face reliance on software from large corporations, which may shape societal development, with few alternatives available.

Opportunities

The National Library has digitized a substantial part of Norway's cultural heritage for training large language models.
The collection is extensive, meeting high standards for training data size.
The library's metadata facilitates addressing issues of representativeness and diversity.
Recently, the library has focused on creating training data from its collection, developing foundational models, and fine-tuning models for specific tasks.

We can convert our digital collection into training data for both generative and non-generative AI models.

7 of 14

How do the National Library ensure that training data is reliable?

Materials are sourced directly from publishers, such as publishing houses, media outlets, and the Norwegian Broadcasting Corporation.
The digitization is performed by their team using specialized equipment and digital solutions.
NB stores the digital content in their data center, keeping three copies across two different technologies and locations, along with digital fingerprints to track file changes.
As a library, they have extensive knowledge of the collection, including both content and technical aspects.

They argue that there is greater trust in these AI models than in those created from a random mix of Internet data.

8 of 14

Norway is in a unique position

With the digital collection at the National Library, Norway is in a unique position to leverage the advancements made in AI in recent years.
In the 2025 state budget, the government allocates 20 million to the National Library to establish a unit for training, updating, and making Norwegian and Sami language models available. The government has deemed it crucial to train such models for language, cultural, and knowledge policy reasons.
The background for the National Library's mission to train Norwegian language models is to ensure that there are models that both public administration and commercial actors can use, based on Norwegian language, Norwegian culture, and knowledge about Norway and unique Norwegian conditions.
Recent international political developments show that there are also central security policy reasons for training Norwegian language models.

Thanks to Wilfred Østgulen, IT Director at the National Library for lending his information to make this presentation.

9 of 14

Collaborative AI Insights in Public Libraries

10 of 14

Background

Agder County Library, in collaboration with librarians, users, service designers, and technologists, has conducted a co-creation project to explore how artificial intelligence (AI) can create value in the library sector. Conducted in autumn 2024, the aim was to identify concrete opportunities to improve services, streamline workflows, and make the library more accessible.

Method

The project followed a user-centered approach, using interviews, surveys, and a full-day workshop. More than 200 people contributed, including 97 librarians and 132 users.

�

11 of 14

Key Findings

Needs and Challenges

Users want personalized recommendations, better search, and easier access to digital resources.
Staff need tools for content creation and wish to spend less time on routine tasks.
There is a need for improved language support, especially for minority language speakers.
Key challenges include fragmented systems, manual processes, low digital competence, and concerns related to ethics and data privacy.
Technological solutions must not replace the human dimension, which both the audience and employees appreciate.

12 of 14

Proposed AI-based Solutions

Enhanced search – Using natural language understanding and semantic search.
Personalized book recommendations – Based on loan history and preferences.
Support for content creation – Efficient tools for thematic displays and communication.
Data-driven collection development – Analytics and prediction to optimize resources.
Technical assistance – Real-time guidance from AI-powered helpers.
Personal AI assistant – Digital guidance and recommendations 24/7.
Early reading support – Interactive tools and parental guidance.
Language support and translation – Real-time multilingual communication tools.

13 of 14

Recommendations for Next Steps

Verify user needs and define success criteria.�Which ideas have the greatest value for users?
Conduct functional and technical evaluations.�Consider complexity, data volume, and privacy requirements.
Start with early pilots and systematic staff training. �The use of AI in digital services is still in its early stages. It is recommended to conduct early and systematic user testing to identify potential challenges. Library staff need thorough training to understand how new solutions work and how results are produced.
Implement in a flexible way with continuous evaluation.

Thanks to Leon Bang-Hetlevik, Advisor at Agder County Library, for the information on the project.

1 of 14