Are Algorithm Registers Transparent? Perspectives from Germany

Existing German AI registers fall short of transparency goals: The audit of Germany's two main AI transparency initiatives (Marktplatz der KI-Möglichkeiten and Platform Lernende Systeme) found that "several adaptations are likely needed for these registers to serve as a useful transparency instrument," indicating current platforms do not adequately fulfill their intended transparency objectives.

Germany lacks coordinated national oversight: Rather than a single federal register, Germany has a "notably fragmented" landscape with "at least five state- and federal-level initiatives" publishing information about AI systems with "varying scopes and objectives," creating confusion about where to find reliable information about government AI use.

Transparency registers need clear technical and governance standards: The paper identifies specific "technical and governance requirements" necessary for effective AI transparency registers, suggesting that without standardized criteria, registers cannot meaningfully inform the public about how algorithms are being used in public administration.

Practical audit tools can drive improvement: By converting conceptual transparency goals into "structured checklists," the researchers created an actionable framework that practitioners can use to evaluate and improve existing registers, demonstrating that transparency gaps can be systematically identified and addressed.

Transparency levels should be visualized and communicated clearly: The paper proposes "a visualization of register transparency levels," suggesting that making it easy for non-experts to understand how transparent (or opaque) different AI systems are could help citizens and policymakers hold government accountable.