Signal A Global Comparison of Schemas, Transparency, and Interoperability in Public-Sector AI Registers and Inventories

Summary

A study conducted an international comparison of AI registers and inventories used to make government AI use visible. It examined 8,368 records drawn from country-level and transnational inventories spanning 72 countries, organized around 23 shared fields. The comparison found that while these registers commonly captured basic descriptive information, they rarely asked for details on appeals processes, risk assessments, legal grounds, or external evaluation. Registers with broader field sets often still left many entries blank, and there was no clear trend toward converging schema structures across sources. Even multiple registers covering the same countries overlapped only partially in content. Based on these patterns, the researchers proposed a layered framework for understanding register visibility as a product of disclosure arrangements, arguing that true interoperability requires shared concepts, clear definitions, and preserved provenance information.