공정거래법상 알고리즘 담합 책임의 귀속과 준법사항(compliance)에 관한 연구

Abstract

Recently the competition law enforcement against algorithmic collusion has been drawn attentions. The possibility of algorithmic collusion and the necessity of competition law enforcement against such algorithmic collusion are generally accepted. Therefore whether or not an undertaking should be responsible for algorithmic collusion is the important matter. This matter should be reviewed with being categorized into algorithmic collusion an undertaking is involved in and algorithmic collusion an undertaking is not involved in. Considering a joint study of the French competition authority and the German cartel office algorithmic collusion an undertaking is involved in includes (i) collusion using algorithms as supporters or facilitators of traditional anti-competitive practices, and (ii) algorithm-driven collusion between competitors involving a third party with an undertaking’s knowing its competitors use the same or coordinated third party algorithms. In case that algorithmic collusion an undertaking is involved in algorithms is supporters or facilitators of collusion and such collusion happens between undertakings. Therefore there is the meeting of minds required under the Korean competition law and so it would not be difficult for an undertaking to be liable for algorithmic collusion. In case that algorithm-driven collusion between competitors involving a third party with an undertaking’s knowing its competitors use the same or coordinated third party algorithms it would be reviewed as concerted practices, conscious parallelism, and information exchange. From the perspective of the Korean competition law this kind of algorithmic would be regarded as a means of information exchange and so the collusion between the undertakings would be accepted with the plus factors required under the Korean precedents. Algorithmic collusion an undertaking is not involved in is the case that such collusion was induced by the parallel use of individual algorithms without communications between the undertakings. This is the collusion induced by black-box algorithms which has the high graded autonomy without an undertaking’s involvement. This case looks that there would not be an undertaking’s accountability for algorithmic collusion because there is no an undertaking’s involvement. Under the Korean laws algorithms is neither a person nor a juristic person and so algorithms cannot be subject of rights and duties under the Korean civil law. Under the Korean criminal law algorithms is not a person and so algorithms cannot commit crimes. Also, because the Korean competition law applies to an undertaking algorithms cannot be liable for collusion. Therefore under the current Korean laws an undertaking shall be liable for algorithmic collusion which such undertaking is not involved in. An undertaking has the obligation not to collude under the Korean competition law. This obligation includes the preventive obligation that an undertaking shall prevent his own collusion actively. With considering an undertaking’s preventive obligation under the Korean competition law algorithmic collusion which an undertaking is not involved in shall be reviewed based on whether or not the undertaking took the preventive measures. In case that an undertaking did not take any preventive measures at all the undertaking should be liable for the result of algorithmic collusion. In case that algorithmic collusion takes place despite the preventive measures such algorithmic collusion means mainly the conscious parallelism, quasi-concerted practices, and information exchange. In such cases the plus factors required by the court should be considered and then the undertaking shall be liable for such practices. From an undertaking’s viewpoint compliance including the preventive measures in the design phase of algorithm and the preventive measures in the implementation phase of algorithm is important in order that an undertaking will perform its preventive obligation under the Korean competition law effectively. Compliance of an undertaking can diminish the possibility of algorithmic collusion in advance and can be an effective means that limits an undertaking’s accountability for algorithmic collusion.