A Conceptual Crosswalk Between Trustworthy Records and Trustworthy AI

Abstract Trust is fundamental to the functioning of societies, yet the rise of artificial intelligence (AI) has intensified challenges of misinformation, opacity, and accountability. This paper examines the concept of trustworthiness through a comparative lens, exploring how it is conceptualized in archival science and in emerging frameworks of Trustworthy Artificial Intelligence (TAI). Drawing on a conceptual crosswalk alongside a scoping literature review, the study identifies conceptual parallels and divergences between the two domains. Archival science grounds trustworthiness in the enduring qualities of records – authenticity, reliability, integrity, and usability – ensuring their evidential and societal value over time. TAI, by contrast, emphasizes transparency, explainability, fairness, and reliability at the system level. Based on a scoping review of 151 papers on Trustworthy AI, this study shows that the literature predominantly foregrounds characteristics such as transparency (76.3%), explainability (68.6%), and fairness (66.7%), while 98.7% of papers provide some form of definition but only 29.9% offer an explicit one. Through a conceptual crosswalk with archival attributes of authenticity, reliability, integrity, and usability, the analysis demonstrates that integrating archival perspectives reframes trustworthiness from a predominantly technical checklist into an evidence-based data governance function grounded in provenance, documentation, and lifecycle accountability within TAI frameworks. The analysis suggests that a cross-domain perspective can deepen TAI’s conceptual foundations by foregrounding the trustworthiness of records – integrating provenance, documentation, and accountability into data governance frameworks – and positioning trustworthiness as an ongoing institutional and ethical commitment rather than a transient technical attribute.