Cai Jingmeng， Journal of China University of Political Science and Law (CSSCI Expanded)

Abstract: Collusion in the field of artificial intelligence (AI) is normally shown in the form of monopoly agreements. Identifying and analyzing the monopoly agreements in the AI field is one of the greatest challenges in the antimonopoly area. On the one hand, assisted by AI, collusion becomes easier and more secret. On the other hand, due to the complexity and unpredictability of AI, companies have strong incentive to cooperate. These can together lead to enforcers’ mis-judgement, resulting in false positive or false negative in anti-monopoly enforcement. The article adopts a “risk-based” classification approach to categorize agreements in the AI field into three tiers. Based on the classification, it points that the factor of “consensus”should be de-emphasized, while “foreseeability” should be prioritized as a key factor. As to high-risk agreements, they should be regarded as “by object”violation, which can be exempted under extremely strict conditions. The medium-risk agreements may include R&D agreements and standard-setting agreements, which could be subjected to block exemption and the analyzing pattern of “burden of proof shifting”. The low-risk agreements shall be generally legal under the Anti-monopoly Law . Finally, the article suggests adjusting the object of the Anti-monopoly Law to face the unsolved problems and novel challenges in the AI area.

Key words: artificial intelligence, monopoly agreement, Anti-monopoly Law, analysis of competitive risk