Top 10 FAIR Data Things for AI

This booklet presents a concise, practice-oriented version of the content developed in the Skills4EOSC D6.3 deliverable titled “Top 10 FAIR Data Things for AI and Health Technology”. It focuses on the Artificial Intelligence section of that document.
It is intended as a quick-access resource for researchers, data professionals, and developers working to apply FAIR principles (Findable, Accessible, Interoperable, Reusable) in Artificial Intelligence workflows, especially those involving machine learning models and datasets.
The 10 practices highlighted here were identified and validated through a collaborative, two-round Delphi study involving domain experts in ML/AI and FAIR. The aim is to share easy, practical steps to help make machine learning and AI models more FAIR. The materials of the Delphi study are available on Zenodo (link: https://zenodo.org/ records/16536643).
In addition to the practices themselves, this booklet briefly summarises key reflections that emerged during a final community discussion, to provide context and indicate future directions.
The work focuses on machine learning (ML) models because they are widely used across disciplines and offer a practical entry point for developing FAIR implementation guidelines. This focus also enabled collaboration with ongoing, well-aligned initiatives. While the study did not distinguish between types of ML (e.g., supervised or reinforcement learning) or delve into more complex models like deep learning, the approach and outcomes are intended to be broadly applicable across different AI model types.