This paper examines the current landscape of AI regulations across various jurisdictions, highlighting divergent approaches being taken, and proposes an alternative contextual, coherent, and commensurable (3C) framework to bridge the global divide. While the U.N. is developing an international AI governance framework and the G7 has endorsed a risk-based approach, there is no consensus on their details. The EU, Canada, and Brazil (and potentially South Korea) follow a horizontal or lateral approach that postulates the homogeneity of AI, seeks to identify common causes of harm, and demands uniform human interventions. In contrast, the U.S., the U.K., Israel, and Switzerland (and potentially China) have pursued a context-specific or modular approach, tailoring regulations to the specific use cases of AI systems. Horizonal approaches like the EU AI Act do not guarantee sufficient levels of proportionality and foreseeability; rather, this approach imposes a one-size-fits-all bundle of regulations on any high-risk AI, when feasible, to differentiate between various AI models and legislate them individually. The context-specific approach holds greater promise, but requires further development regarding details, coherent regulatory objectives, and commensurable standards. To strike a balance, this paper proposes a hybrid 3C framework. To ensure contextuality, the framework bifurcates the AI life cycle into two phases: learning and utilization for specific tasks; and categorizes these tasks based on their application and interaction with humans as follows: autonomous, discriminative (allocative, punitive, and cognitive), and generative AI. To ensure coherency, each category is assigned regulatory objectives. To ensure commensurability, the framework promotes the adoption of international industry standards that convert principles into quantifiable metrics to be readily integrated into AI systems.

翻译：本文审视了不同司法管辖区当前的人工智能监管格局，凸显了各国所采取的分歧性路径，并提出了一种替代性的情境化、连贯且可通约的（3C）框架，以弥合全球分化的鸿沟。尽管联合国正在制定国际人工智能治理框架，七国集团也已认可基于风险的路径，但各方在其具体细节上仍未达成共识。欧盟、加拿大和巴西（以及潜在韩国的）遵循横向或平行路径，这种路径假定人工智能的同质性，试图识别共同的损害原因，并要求统一的人为干预。相比之下，美国、英国、以色列和瑞士（以及潜在中国）则采取了情境特定或模块化路径，针对人工智能系统的具体使用场景量身定制监管措施。诸如《欧盟人工智能法案》之类的横向路径未能确保充分的相称性和可预见性；相反，这种路径对任何高风险人工智能施加了一套"一刀切"式的监管措施，而在可行时则应对各类人工智能模型加以区分并分别立法。情境特定路径更具前景，但在细节完善、监管目标连贯性以及可通约标准方面仍需进一步发展。为取得平衡，本文提出了一种混合型3C框架。为确保情境性，该框架将人工智能生命周期分为两个阶段：特定任务的学习与利用；并根据应用场景及与人类的交互方式将这些任务分类如下：自主型、判别型（分配型、惩罚型与认知型）及生成型人工智能。为确保连贯性，每个类别均被赋予相应的监管目标。为确保可通约性，该框架倡导采纳将原则转化为可量化指标并可直接集成至人工智能系统的国际行业标准。