From millions of documents to searchable insights in six days

When several million pages of the Epstein documents were made public, TV 2 quickly needed a tool that would enable journalists to search, analyze, and find relevant connections in enormous amounts of unstructured information.

The challenge

In January 2026, about 3.5 million pages of the Epstein documents were published. Newsrooms around the world immediately began searching for new stories, famous names, and previously unknown information.

The documents were unstructured and had to be downloaded and made available before they could be analyzed. TV 2 needed a system that could handle large amounts of data and make the content searchable, and they needed it in a very short time.

Solving

In just two days, we developed the first version of a lookup tool for the Epstein documents. Four days later, the solution was in production and being used by journalists.

The project was carried out in close collaboration between developers and journalists, with continuous feedback and rapid iterations. Because the material contained sensitive documents, access control was built in from the start.

Artificial intelligence also played a central role in the solution. Image analysis was used for facial recognition and automatic image annotation, so that visual content could also be made searchable and analyzed together with the rest of the document material.

Value

When the news landscape changes from hour to hour, speed of development is crucial. The solution gave TV 2 a tool that enabled journalists to work effectively with extensive and confusing source material, and helped the editorial team quickly identify and investigate new cases.

The architecture and working methods from the project can also be reused in future document leaks or other excavation projects with large amounts of unstructured data.