GAI算法垄断风险研究——基于法制规范的视角

Risks of Algorithmic Monopoly in Generative Artificial Intelligence—A Study from the Perspective of Legal Regulation

摘要: 在深度学习与大模型技术的推动下，生成式人工智能正迅速嵌入市场运行的核心结构。其在数据处理、内容生成与决策优化等方面所展现出的能力，使算法由传统的经营辅助工具逐步转变为深刻影响竞争秩序的制度性力量。相较于一般算法，生成式人工智能在自主学习、系统不透明性以及规模依赖性等方面呈现出更为突出的特征，客观上强化了平台经营者的市场支配地位，并诱发算法合谋、自我优待以及竞争排除等新型垄断风险。现行反垄断法以既成行为与明确竞争损害为中心构建分析框架，在生成式人工智能语境下面临市场界定困难、垄断行为认定失灵以及责任归属模糊等制度性困境。本文在系统分析生成式人工智能算法垄断风险的生成机理及其主要表现形态的基础上指出，反垄断法在应对该类风险时所面临的核心问题，并非规范强度不足，而在于既有分析工具与规制理念未能有效匹配生成式人工智能的技术特征。为此，有必要在坚持反垄断法基本价值的前提下，引入风险导向的规制思路，对相关规则作出有限而关键的适应性调整，以回应生成式人工智能对竞争秩序带来的结构性冲击。

Abstract: With the rapid advancement of deep learning and large-scale model technologies, generative artificial intelligence (GAI) has become deeply embedded in the core structures of market operation. Its growing capacity in data processing, content generation, and decision-making optimization has transformed algorithms from auxiliary managerial tools into institutional forces that profoundly shape competitive order. Compared with traditional algorithms, generative artificial intelligence exhibits more pronounced characteristics of autonomous learning, systemic opacity, and scale dependence. These features objectively reinforce the market power of dominant platforms and give rise to new forms of algorithmic monopoly risks, including tacit collusion, self-preferencing, and exclusionary practices. The existing antitrust framework, which is primarily built upon the ex post identification of concrete conduct and demonstrable competitive harm, encounters substantial institutional difficulties when applied to generative artificial intelligence. Challenges arise in market definition, conduct assessment, and the attribution of legal responsibility. This article argues that the core problem in regulating algorithmic monopoly risks associated with generative artificial intelligence does not lie in insufficient normative intensity, but rather in the structural mismatch between traditional antitrust analytical tools and the technological characteristics of generative artificial intelligence. Accordingly, while adhering to the fundamental values of competition law, it is necessary to introduce a risk-oriented regulatory approach and to implement limited yet critical adaptive adjustments to existing rules, so as to effectively address the structural impact of generative artificial intelligence on competitive order.