Distinguishing Task-Specific and General-Purpose AI in Regulation

TL;DR

This paper addresses regulatory challenges posed by general-purpose AI (GPAI), highlighting four distinct features that set GPAI apart from task-specific AI: generality, evaluation difficulty, new legal concerns, and a distributed value chain. It argues that policy should focus on applications and outcomes rather than model specifications, and introduces a risk-chain framework to analyze how harms propagate across actors and stages of deployment. The authors propose three concrete recommendations: regulate by use-case and impact; use technical proxies only in narrow, justified circumstances; and strengthen structural constraints across the entire risk chain. By outlining these distinctions and strategies, the work provides policymakers with a practical roadmap to govern GPAI more effectively and avoid misapplied or outdated regimes.

Abstract

Over the past decade, policymakers have developed a set of regulatory tools to ensure AI development aligns with key societal goals. Many of these tools were initially developed in response to concerns with task-specific AI and therefore encode certain assumptions about the nature of AI systems and the utility of certain regulatory approaches. With the advent of general-purpose AI (GPAI), however, some of these assumptions no longer hold, even as policymakers attempt to maintain a single regulatory target that covers both types of AI. In this paper, we identify four distinct aspects of GPAI that call for meaningfully different policy responses. These are the generality and adaptability of GPAI that make it a poor regulatory target, the difficulty of designing effective evaluations, new legal concerns that change the ecosystem of stakeholders and sources of expertise, and the distributed structure of the GPAI value chain. In light of these distinctions, policymakers will need to evaluate where the past decade of policy work remains relevant and where new policies, designed to address the unique risks posed by GPAI, are necessary. We outline three recommendations for policymakers to more effectively identify regulatory targets and leverage constraints across the broader ecosystem to govern GPAI.