How Data-Related AI Research can Support Technical Solutions for Regulatory Compliance

Danilo Brajovic, David A. Kreplin, Marco Huber

Keywords: Regulation, Data, AI Act

TL;DR: We show how current research on ML data can support ML regulation

Abstract: Ensuring high-quality, representative, and secure datasets is critical for compliance with emerging regulatory frameworks such as the EU AI Act (Art. 10). In this paper, we survey five key data-centric challenges: intrinsic and context-dependent data quality, availability, variability, and security, and link each challenge to established and emerging methods and research. We then propose a workflow that integrates best practices from machine learning research with regulatory requirements, illustrating how each step can be operationalized to meet the “relevant, representative, error-free” criteria. Our analysis highlights opportunities for regulators to refine their mandates by incorporating advances in ML research.

Submission Number: 29