The black box challenge: Legal perspectives on explainability and algorithmic transparency

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

This study addresses the regulatory crisis triggered by the "black box" challenge in artificial intelligence, testing the hypothesis that algorithmic opacity represents an extension of risk theory that upends classical liability and causal attribution. Methodologically pairing a technical assessment of machine-learning architectures with a functional, comparative analysis of the EU AI Act and the GDPR, the paper exposes a core algorithmic paradox: the computational complexity driving AI’s predictive power simultaneously dismantles traditional mechanisms of legal accountability. Ultimately, the study demonstrates that while statutory explainability mandates are indispensable for securing procedural due process, they remain structurally bounded by intellectual property rights and post-hoc technical instability. Consequently, a viable regulatory regime must transcend localized explanation rights, supplementing them with a composite network of independent auditing, ex-ante risk assessments, and robust judicial review calibrated through a flexible, risk-based proportionality matrix.