[Paper Review] A Comprehensive Review and Systematic Analysis of Artificial Intelligence Regulation Policies

Weiyue Wu, Shaoshan Liu|arXiv (Cornell University)|Jul 23, 2023

Ethics and Social Impacts of AISocial Sciences3 citations

TL;DR

This paper presents a systematic framework for analyzing global AI regulation policies by categorizing them through rationale and approach, identifying systemic flaws in risk-based and performance-based models, and offering a structured, divide-and-conquer method to help governments develop balanced, adaptive, and effective AI regulations grounded in historical lessons and contextual analysis.

ABSTRACT

Due to the cultural and governance differences of countries around the world, there currently exists a wide spectrum of AI regulation policy proposals that have created a chaos in the global AI regulatory space. Properly regulating AI technologies is extremely challenging, as it requires a delicate balance between legal restrictions and technological developments. In this article, we first present a comprehensive review of AI regulation proposals from different geographical locations and cultural backgrounds. Then, drawing from historical lessons, we develop a framework to facilitate a thorough analysis of AI regulation proposals. Finally, we perform a systematic analysis of these AI regulation proposals to understand how each proposal may fail. This study, containing historical lessons and analysis methods, aims to help governing bodies untangling the AI regulatory chaos through a divide-and-conquer manner.

Motivation & Objective

• To address the global chaos in AI regulation by providing a comprehensive review of proposals from diverse geopolitical and cultural contexts.
• To develop a unified analytical framework that categorizes AI regulation proposals based on rationale and approach for systematic evaluation.
• To identify structural and operational weaknesses in existing AI regulation models, particularly risk-based and performance-based approaches.
• To guide governing bodies in evolving AI regulations through evidence-based, iterative policy refinement using historical analogies.
• To support the long-term goal of achieving a globally coordinated, standardized AI regulatory framework.

Proposed method

• Conduct a comparative review of AI regulation proposals from the U.S., EU, UK, China, and international bodies, focusing on their underlying rationales and implementation approaches.
• Develop a two-dimensional analytical framework using 'rationale' (regulatory focus) and 'approach' (implementation mechanism) to classify and evaluate regulatory models.
• Apply the framework to systematically assess the strengths and failure modes of each regulation model, especially in dynamic and uncertain AI environments.
• Draw analogies from historical regulatory domains—such as privacy, cybersecurity, and consumer protection—to inform AI regulation design and avoid past pitfalls.
• Identify key vulnerabilities in risk-based systems (e.g., ambiguous categorization) and performance-based systems (e.g., lack of standards, outcome unpredictability).
• Propose a 'divide-and-conquer' strategy for policy evolution, enabling iterative, context-specific, and evidence-driven regulatory development.

Experimental results

Research questions

• RQ1 How do regulatory rationales and approaches differ across major AI regulatory frameworks such as those in the U.S., EU, UK, China, and international proposals?
• RQ2 What are the inherent failure modes of risk-based AI regulation, particularly regarding dynamic risk categorization and enforcement ambiguity?
• RQ3 Why might performance-based regulation lead to catastrophic outcomes in AI, especially in the absence of standardized benchmarks or ethical guardrails?
• RQ4 How can historical regulatory models (e.g., in privacy, cybersecurity, and consumer protection) inform the design of more effective and adaptive AI regulations?
• RQ5 What systematic method can help governing bodies untangle the current regulatory chaos and evolve toward a globally coherent AI governance framework?

Key findings

• The EU’s risk-based approach, while aiming for flexibility, risks regulatory ambiguity due to inconsistent and shifting risk categorizations across AI applications.
• Performance-based regulation in the UK and proposed international AI agencies may lead to catastrophic outcomes due to insufficient standards and unchecked developer autonomy.
• The U.S. industry-specific model, though effective in domains like privacy and cybersecurity, lacks a unified horizontal AI regulatory strategy, leading to regulatory fragmentation.
• China’s top-down, state-led regulatory model emphasizes control and national security, but may hinder innovation and international interoperability.
• The absence of universally accepted benchmarks or performance metrics undermines the feasibility of outcome-focused regulation, especially in high-stakes AI applications.
• A systematic, context-aware framework that combines historical analogies and structured categorization is essential to prevent regulatory myopia and enable adaptive, long-term AI governance.