Data discovery — a tax incentive for sharing data

Data is the oil of the new KI economy. Without data, you can't train KI and you can't use it either. If we manage to produce, cultivate and share data, we can create an economic revolution. The genius of data is that it can be used by everyone at the same time and the more people who use and refine the data, the more value it gets.

Our problem is not a lack of data. The Norwegian business sector has large amounts of data. The problem is that they are not applied across businesses. The data is often unstructured, it is costly to clear and document it, and the business doing the job gets no financial gain. Then data sharing will naturally be downplayed, although the societal benefit could be great.

This is a classic coordination problem. Each individual actor bears the cost of making data usable, while the gain comes only when many actors do the same. Today, this coordination is absent.

KI relies on large amounts of structured data. Without high-quality data, there are no models that can be trained, no automation and no productivity gains. According to the Stanford AI Index 2024, a significant portion of global AI investment goes to infrastructure -- not just algorithms. It shows that computing power, storage and systems are becoming as fundamental to value creation as the models themselves.

Norwegian companies have plenty of domain expertise and data sources, but lack an economic system that makes it profitable to elevate this data to a level where it can actually be used by KI — in their own business or across value chains.

Tax findings provide a model for how a tax incentive can lower the threshold for desired activity. The scheme is hotly debated, but it illustrates one good mechanism: when companies face coordination problems, in this case research and development work that everyone benefits from, not just the company that pays for it, a tax credit for documented effort can provide more of what we want and distribute the cost of research and innovation more equitably. The technology area of data lacks such a means.

A Datafunt scheme can use the same logic. Companies get tax credits when they elevate their own data sets to a level where they can actually be used by others in a value chain. Requirements must be easy for businesses to understand, and be about practical data quality — not advanced technology:

• The dataset must be described so that others can understand what it contains (metadata, source, time period, device).
• It must be structured so that it can be used without manual cleanup jobs (for example, same column layout, same concepts, and same units of measurement).
• It must be possible to access the data under a defined regime, either openly, partnership-based or through a paid license.

Why does this matter? Because shareable data reduces friction throughout the economy. In energy, transport, construction, aquaculture and health, there are often multiple actors that rely on compatible data sets. Once data is structured equally and described equally, integration costs fall rapidly, and the use of automation and AI becomes possible. The company that does the job gets gains internally, too: better management data, less manual reporting and more robust processes.

Internationally, there are several examples of governments rewarding work with data quality and sharing. In the UK, the new Data (Use and Access) Act 2025 authorities are responsible for establishing mechanisms that make data sharing easier and more cost-effective, in particular through standardised access models in finance, energy and transport. In Singapore, IMDA ensures that companies can share data safely and commercially, through a national framework - Trusted Data-Sharing Framework — enabling standardised data partnerships, validating datasets for reuse, and testing data-driven innovations in regulated “sandboxes” before launch. In the European Union, demands Data Governance Act that member states establish mechanisms to fund data sharing in key sectors, in particular through so-called “data altruism organisations” and certified sharing mechanisms.

Norway lacks a similar instrument. We have good common components -- Altinn, health data directories, financial industry API standards, but we don't support the costly work that needs to be done before data can be shared. That job -- clearing, standardizing and describing the data sets -- falls today between all trusts, because no one gets paid for it.

Tax incentives won't solve everything. But when the market lacks economic motives for producing a common asset — in this case, data quality and compatibility — they are among the most accurate means. A Datafund scheme will make data sharing an investment, not a cost, and provide Norwegian business with a foundation for AI and automation that would otherwise take many years to build.