This comment refers to the paper in its entirety. Most of the paper is devoted to summarizing other sources that deal with various aspects of AI assurance. Topics range widely from discussions of AI models to trustworthiness, cybersecurity, project lifecycles,  governance, risks, to name a few. Rarely, however, does the paper deal with the topics at hand: how to get the most out of the available data, how to document the sources and manipulations that data, how to document the limitations of the data sets, and how to integrate data from different sources for use in machine learning. There are extensive discussions of the challenges and desired outcomes, but little or no explanation of what should be done on any practical level. I would encourage the authors to think about what is needed to fill the gaps, and in particular, what role HL7 should play in addressing those gaps.