How Supply Chain Dependencies Complicate Bias Measurement and Accountability Attribution in AI Hiring Applications

Bias emerges from how AI components interact, not from individual components alone. The paper demonstrates that a resume parser functioning without bias independently can still contribute to discrimination when combined with specific ranking algorithms and filtering thresholds—meaning companies cannot simply audit individual vendors' tools to ensure fairness.

Current regulatory frameworks don't address fragmented responsibility across supply chains. While the EU AI Act, NYC Local Law 144, and Colorado's AI Act attempt to regulate AI hiring systems, they overlook the fundamental problem that responsibility fragments across data vendors, model developers, platform providers, and deploying organizations, leaving accountability gaps.

Information asymmetries create a compliance illusion where everyone appears compliant but the system discriminates. Deploying organizations bear legal responsibility without technical visibility into vendor-supplied algorithms, while vendors control implementations without meaningful disclosure requirements. The paper notes that "each stakeholder may believe they are compliant; nevertheless, the integrated system may produce biased outcomes."

Proprietary configurations prevent integrated bias evaluation. Because vendors keep their configurations proprietary, organizations cannot conduct comprehensive bias testing across the entire hiring system—they can only evaluate isolated components, missing the discrimination that occurs at integration points.