Most AI governance instruments concentrate on whether a system is sound — accurate, risk-managed, compliant. The major frameworks — the NIST AI Risk Management Framework, the EU AI Act, ISO/IEC 42001 — answer that with increasing rigour, and the EU AI Act's Fundamental Rights Impact Assessment begins to reach past it toward the people a system affects. But a second question remains largely unaddressed, and it is the one the people on the receiving end of these systems most need answered: is this system trustworthy to those it affects — does it build their capacity rather than deplete it, distribute power rather than concentrate it, and answer to them rather than only to regulators? This paper presents the AI Trust Infrastructure Assessment Framework, a diagnostic built for that second question. It assesses any AI deployment across five dimensions — epistemic integrity, capacity orientation, power architecture, contextual intelligence, and accountability architecture — through fifteen criteria scored on a five-point maturity scale. Its defining move is to score the same criteria from two independent vantage points: the builder, scoring what was designed into the system, and the adopter, scoring what happens when the system meets real people, processes, and context. The divergence between those scores — the Alignment Gap — is the diagnostic signal. Where builder and adopter diverge by two or more levels on any criterion, governance is failing at a specific, locatable point, usually before harm becomes visible in outcome metrics. The framework is designed to complement existing technical standards, not replace them: it occupies the governance, power, and capacity layer those standards were not built to cover. The full scoring rubric, two worked case studies, and notes on use and limitation are included. Working paper, Version 1.0. Submitted to AI and Ethics (Springer Nature). This is the author's version.
Sonali Sharma (Fri,) studied this question.