Sweden’s government on Friday announced that it was developing its own Swedish-language model for artificial intelligence.
“The government has adopted a Swedish AI strategy, and an important part of it is that Sweden needs high-quality AI in Swedish,” Prime Minister Ulf Kristersson told a press conference.
“For that, a high-quality Swedish-language model is required,” he added.
Such “large language models” (LLM) power chatbots like OpenAI’s ChatGPT and Google Gemini. The systems are trained on data in multiple languages.
“Language models are not just translated words, they also carry history, culture, traditions, values, everything that is embedded in a language,” Kristersson said.
When used “they shape how information is interpreted, prioritised, and communicated,” he continued, adding that therefore a homegrown model was a “strategic ability” for the development of AI in Sweden.
He said representatives from the business community, authors and publishers, media companies, research institutions and interest groups had been gathered to collaborate on Swedish AI.
Sara Mazur, executive of the Knut and Alice Wallenberg Foundation — which funds the Wallenberg AI, Autonomous Systems and Software Program (WASP) launched in 2015 — said the model would be based on ones that are already available.
She added that work would begin immediately and they hoped to complete most of the “training” of the model during 2026.
Mazur said authors, publishers and news media in Sweden had agreed to contribute “high-quality, editorially reviewed training data that reflect Swedish values and Swedish norms”.
“The goal is to create a language model that not only speaks and writes Swedish, but also understands the Swedish context and can communicate on that basis,” Mazur told reporters.
Add Comment