Signal Paper Argues AI Regulation Must Address Unreliable Inference, Not Just Data Protection

Summary

This paper, by A Mukundan, Debayan Gupta and Subhashis Banerjee, starts from the observation that AI systems increasingly mediate decisions affecting individuals and societies. The authors acknowledge that existing data protection frameworks do address certain privacy harms, such as those arising from data leakage, re-identification and profiling. They argue, however, that these frameworks fail to capture a more fundamental risk, namely unreliable or unjustified inference produced by AI systems even when the underlying data collection is lawful and consensual. The authors therefore propose treating the validity, reliability and transparency of inference as a distinct regulatory concern, separate from conventional data-protection obligations. They argue that AI governance frameworks need to evolve beyond data-centric privacy law to address this gap.