DLT as an Audit and Trust Layer for AI Decisions

A Practice-Based, Architecture- and Governance-Oriented Classification

Abstract

In recent years, artificial intelligence has become a central component of digital transformation strategies and is increasingly influencing operational decision-making processes within organisations. Advances in machine learning have led to AI systems being increasingly integrated into operational decision-making processes. At the same time, however, it is becoming apparent that many AI systems are only able to manage the transition from pilot applications to audit-proof and accountable production environments to a limited extent.
In practice, AI decisions are often embedded within complex data and system landscapes that have evolved over time. Unclear responsibilities and a lack of traceability regarding data states, model versions and decision-making contexts hinder the trustworthy deployment of AI. These shortcomings relate less to the performance of the models than to the structural requirements for their governance.
This paper presents a practice-oriented architectural concept that addresses the need for sustainable trust in AI decisions by introducing an additional audit and trust layer. Against this backdrop, it demonstrates how distributed ledger technologies (DLT) can be utilised as an independent and immutable infrastructure for logging key AI artefacts. The aim is to present DLT as a complementary building block for governance, auditability and institutional accountability in AI system architectures.