“INN-the-Loop” was a research project on artificial intelligence (AI) for critical sectors in 2024 that analysed limitations and risks related to ML, GenAI and Agentic AI. It identified significant risks and barriers, as well as solutions for adoption of AI in healthcare and other critical sectors that demand high standards of accuracy, safety, security, and privacy.
The project was managed by the University of Inland Norway (INN) with contributions from the Norwegian Computing Center (NR), SINTEF, NTNU SFI NORCICS (Norwegian Center for Cybersecurity in Critical Sectors), and VentureNet AS. The project received co-financing by the Regional Research Fund (RFF) Innlandet supported by Innlandet County and the Research Council of Norway.
Abstract
Building on rapid development and investment in Artificial Intelligence (AI), the year 2025 heralded in “Agentic AI” as the new frontier for Generative AI (GenAI). The implication is that virtual assistants will be able to autonomously solve problems, set goals, and increase productivity by automating workflows, generating documents, and enhancing the productivity of humans who use AI-supported systems.
However, for Agentic AI to be suitable for use in critical sectors, a solution is needed to address inherent limitations of AI related to accuracy, safety, security, adaptivity, trustworthiness, and sustainability. This article summarizes results from a research project in 2024 with leading Norwegian research institutions titled the “INN-the-Loop”. The aim of the project was to pre-qualify a framework to design, develop and test human-centric AI-systems for critical sectors, with a focus on smart healthcare as a use case. The project’s findings on AI risks shed light on the importance of digital regulation to ensure safety and security, while also presenting possible solutions for compliance automation to cost-effectively cope with changing regulatory, technical and risk landscapes.
This article describes a framework, methodology and system/toolkit to develop trustworthy and sustainable AI-systems with Humans-In-The-Loop (HITL). The framework aims to address limitations and risks of current AI approaches by combining human-centred design with “Data Space” technologies, including privacy-enhancing technologies (PETs) for decentralised identity and data access management.
The project’s results are aligned with European initiatives to develop federated, sustainable and sovereign digital infrastructure for high performance (HPC) and edge computing. The results can inform design and planning of next-generation digital infrastructure, including local digital twins (LDT) and interconnected digital marketplaces, which can strengthen supply chain resilience in critical sectors.
Download the full research report.
English 
Norwegian