The Deficiency of “Redline/Greenline” Approach to Risk Management in AI Applications

Abstract

Expanding application of technologies classified as AI draws public attention to associated risks and effective measures for risk mitigation. Policy makers in most countries are currently looking for effective approaches to securing the public interest against AI-related risks and unforeseen consequences of widening AI use. In this connection, there is widespread talk of defining “red” and “green” areas for AI technologies, frequently leading to calls to draw “red lines” and “green lines” for technological innovation. The authors draw on the analysis of AI related risks in a number of international fora and question the efficacy of this “redline/greenline” approach in terms of making the benefits of AI available in the society, while not impeding innovation and technological progress. The authors propose that a more nuanced approach is required, which could involve certification of AI system in sensitive applications, or could apply codified ethical principles to derive specific rules for AI use dependent on the risks created by AI in a particular application.