← ALL NEWS

DIGI.NO · 21 Nov 2026 · NORDICS

Norway's AI Needs More Than Good Ambitions

This is an opinion piece from three Accenture Norway leaders: CEO Torbjørn Eik-Nes, head of data and AI Ragnar Wisløff, and head of cybersecurity Martin Sværen. They respond to a busy October in Norwegian AI policy, when the government kicked off a new national AI strategy, signed an AI agreement with the social partners in working life, and presented a budget with a larger AI push. They welcome the pace and ambition and say the government is right to seek input from industry, which has worked on many of these problems for years.

On the budget, they point to the proposed 25 million kroner for a shared generative AI solution in the public sector. Many organisations have already built such solutions with security, access control and privacy in place, so the state should look at what already works before building something new. They cite Accenture's global survey: three in four public sector leaders expect AI to change roles and responsibilities within a year, while six in ten say it is hard to find the needed skills. They argue training should become a fixed part of public sector AI efforts, since a country that cannot understand, evaluate and challenge its technology has limited room to act.

The authors call it unrealistic for Norway to become independent of American or international technology. Their survey found 70 percent of leading AI models in 2025 came from the US, and 65 percent of European businesses say they cannot compete without non-European providers. On average only 36 percent of AI initiatives and datasets require a sovereign solution, so the rest can safely run with global providers. The question, they say, is how much dependence and in which areas.

They welcome the 30 million kroner for AI safety and model evaluation, but want a mapping of which dependencies are strategic. Which public services rely on a single model or cloud provider, and can they switch without stopping? They propose that the strategy set targets for room to act: whether critical services can keep running if a provider drops out, whether alternatives exist where needed, and whether someone can stop an AI agent that misbehaves. The goal is to use the best technology in the world without losing the ability to act when conditions change.

Read the original ↗