NIOM explores local artificial intelligence

Artificial intelligence (AI) is changing how we work across many sectors, and healthcare is no exception. At NIOM, we have over the past six months worked on a project exploring how AI can be used in a safe and responsible way in our professional activities.

A key question has been how we can use AI to improve efficiency without compromising data protection and information security.

Much of the information we work with – including patient-related data – is confidential and cannot be shared with external services or stored on external servers. The approach we are exploring is local AI: models that run directly on our own hardware, without internet access, in a so-called sandbox environment.

Status and next steps

We have established a functional local AI solution and developed functionality based on what is known as Research-Augmented Generation (RAG). This allows the AI model to retrieve relevant information from scientific publications and textbooks, enabling responses grounded in domain knowledge. So far, the system has only been tested on openly available data.

The RAG system has already been used at internal conferences at the University of Oslo, something we are pleased with as a small dental research institute. We expect that collaboration with UiO will support further development going forward.

The project is now established as an ongoing initiative, and we will continue to develop our AI system. The aim is for local AI to support the analysis and handling of data in a way that is efficient, secure, and reliable. We hope this work will contribute to a better understanding of how AI can be used responsibly in healthcare.