CYIL vol. 11 (2020)

CYIL 11 (2020) ARTIFICIAL INTELLIGENCE AND COMPETITION LAW… was originally developed by professors Ariel Ezrachi and Maurice E. Stucke, 49 and it has become a common starting point for further discussions on this topic. Without going into details, we will briefly outline the four basic scenarios: (i) Messenger; (ii) Hub-and-Spoke; (iii) Predictable Agent; and (iv) Artificial Intelligence. 2.1Messenger This scenario involves a “traditional” agreement among humans representing the undertakings, while the algorithms are only used to support or facilitate the agreement, e.g. to implement, monitor, enforce, or conceal the respective anticompetitive practice. 50 This scenario is already known to have been realised in practice. Among the most famous ranks the US Topkins case, 51 also known as the Poster Cartel . David Topkins, the first person in the United States ever to be charged for using an algorithm to fix prices, pleaded guilty for conspiring to fix, increase, sustain, and stabilize prices of posters sold online. A similar case, known according to one of the undertakings involved as Trod , was investigated in the UK. 52 The Messenger scenario is not the domain of only horizontal cartels; in the same way as with horizontal cartels, algorithms may be used to support and facilitate vertical price fixing, as was recently demonstrated in the Commission’s Consumer Electronics cases. 53 From the point of view of imputation of liability, the Messenger scenario does not pose any new questions. The anticompetitive agreement is based on a human agreement and is therefore clearly attributable to the undertakings involved. 54 We will therefore discuss it no further. 2.2 Hub-and-Spoke The concept of Hub-and-Spoke cartels is also not new in competition law doctrine. Whereas, in a “normal” cartel, the cartelists communicate directly with one another, in the hub-and-spoke scenario, they communicate only indirectly, via a third person (hub) who is in contact with all of the undertakings involved (spokes). 55 The role of the hub may be played by an algorithm. From a legal point of view, the situation is relatively straightforward when the undertakings involved knowingly use the algorithm in order to coordinate their prices. This has happened in the Eturas case , in which the CJ EU issued its judgement in 2016 and which concerned a provider of an online travel booking system for travel agencies. The agencies involved were informed that the booking system will, in order to “ normalise the conditions of competition ”, prevent discounts from exceeding 3%. The CJ EU confirmed the position of the Lithuanian 49 EZRACHI, STUCKE ( op. cit. sub 7). 50 Franco-German Study , p. 27. 51 US DOJ press release of 6 April 2015, available at: https://www.justice.gov/opa/pr/former-e-commerce- executive-charged-price-fixing-antitrust-divisions-first-online-marketplace (1 July 2020). 52 TRICKER, M., ROGERS, S. Online Retailers Should Tread Carefully after Trods (2016). Available at: https:// www.nortonrosefulbright.com/en/knowledge/publications/7e7bdcca/online-retailers-should-tread-carefully- after-trod (1 July 2020). 53 EUROPEAN COMMISSION press release of 24 July 2018, available at: https://ec.europa.eu/commission/ presscorner/detail/en/IP_18_4601 (1 July 2020). 54 COLOMBO, N. Human Liability Vis-a-vis Artificial Intelligence’s Anticompetitive Behaviours. Competition and Regulatory Law Review , 2018 (1), pp. 14 & 15. 55 In the EU, decisions on such cartels are more common on national level, see FAULL, J., NIKPAY, A. The EU law of Competition. Third Edition . Oxford University Press, 2014, p. 1051.

247