Algorithms and the use of Artificial Intelligence (AI) have become commonplace in a vast number of markets, and this has drawn the attention not only of competition law academics and practitioners, but also of competition authorities. These authorities have expressed their concerns over the use of these algorithms to engage in anti-competitive practices.

Commissioner Vestager has named “the risk that automated systems could lead to more effective cartels” as one of the future challenges for cartel enforcement.¹

Algorithms as monitoring tools for cartel enforcement

Algorithms can monitor competitors’ behavior in the market, and this can be used as a facilitator of the implementation of cartels and the application of retaliation measures. In this case, humans agree on anti-competitive behavior and subsequently implement the cartel and monitor its implementation using algorithms that can instantly identify deviations and apply retaliation measures.

This case should not present any assessment challenge to competition authorities. There have already been good examples of decisions that show how competition law, as it stands, can deal with algorithms: the Poster and Frames cases in the United Kingdom² and the Topkins case in the United States.³ In these two cases, competing online sellers had agreed not to undercut each other’s prices and implemented this agreement through ‘automated repricing software’ that automatically set the prices of the products so as to stay in line with other online sellers, which restricted price competition between the online retailers on Amazon. The parameters of the anti- competitive conduct seen in these cases are not new. The only novel element is the use of a more sophisticated tool to monitor the correct implementation of the agreement. Anti-competitive agreements are caught by Article 101 Treaty on the Functioning of the European Union (TFEU), irrespective of the instruments used for their implementation.

‘Hub and spoke’ cartels

In these type of cartels, the competing firms use a common third-party algorithm (as the hub) to determine pricing and react to changes in the market, which in turn generates a collusive structure, i.e., the ‘hub and spoke’ cartel. The underlying concern here is that the enhanced use of algorithms will increase the number of possible scenarios where this ‘hub and spoke’ structure can exist, and competitors will trust that, because they all use the same third- party algorithm, the market will lack any uncertainty or competition.⁴

 For the time being, the most similar case to a hub and spoke cartel at the European Union level is the Eturas case.⁵  Following a reference for a preliminary ruling of the Lithuanian Supreme Administrative Court, the Court of Justice of the European Union (CJEU) provided some guidance on the concept of a concerted practice and the elements that should be taken into account to assess whether a concerted practice has been implemented when economic operators use a common computerized information system.⁶

 The Lithuanian Competition Authority found that several online travel agencies had infringed competition law by limiting the possibility of applying discounts of more than 3% when using the Eturas online travel booking system. In turn, Eturas was considered to have played the role of a facilitator in the implementation of the infringement. The Eturas system sent a message to the online travel agencies inviting them to cap the discount rates for travel bookings and informing them that the platform would undergo a technical modification, whereby any discounts in excess of the cap would be automatically reduced. Evidence showed that the message, although available in the Eturas system, was only accessed by two companies, and there had not been any further reaction to the message by the online travel agencies. The restriction in question did not prevent operators from providing higher discounts, but the travel agencies concerned had to take additional technical steps in order to do so.

The Lithuanian Competition Authority had initially considered that the travel agencies: (i) “could reasonably assume that all the other users of that system would also limit their discounts to a maximum of 3%”⁷ and (ii) had showed no objection to the proposed limit of discounts. The CJEU, on the other hand, held that the concept of a concerted practice implies concertation among the companies at hand and consequently their conduct on the market. In this case, if the evidence did not show any awareness, it could not be assumed that the companies participated in a concerted practice, unless there were “other objective and consistent indicia”⁸  showing that the companies tacitly agreed to engage in anti-competitive behavior.

Algorithms as autonomous cartelists

Other theories of harm suggest that the when a company unilaterally employs algorithms to adapt to market changes and to maximize profits, this conduct could lead to tacit collusion restricting competition in the market. These algorithms are self-learning and through their predictive capacity adapt to market conditions and align prices with those of competitors.⁹ For example, pricing algorithms, instead of explicit communication, could be used to signal unilateral pricing intentions.¹⁰ Although unilateral price signaling in the absence of collusion cannot be said to amount to a by object infringement, the European Commission (EC) in the Container Shipping Article 9 commitments decision¹¹  preliminarily found that unilateral public announcements of a series of general rate increases, which are normally not problematic under competition law, allowed companies to signal future pricing intentions leading to an increase of price, because they were non-binding and made long before the client decisions.

According to the EC, such unilateral, non-binding price increase public announcements were of no value to customers and could therefore amount to an infringement of Article 101 TFEU,  even if the EC found no evidence of collusion. Since the commitments decision was adopted under Article 9, it did not allow the EU courts to review whether in the absence of any proof of collusion or awareness, such unilateral price increase announcements could constitute an infringement of Article 101 TFEU.  The EC and other competition authorities will face significant challenges in proving that public price announcements amount to a by object restriction in the absence of collusion, and in applying a similar theory of harm to algorithms and the online world.

Another potential anti-competitive conduct would be that of a company that unilaterally applies algorithms to adapt to market changes and to maximize profits. In this scenario, instead of aligning prices, algorithms would, through self-learning, adapt to other commercial behaviors of competitors when this would be the most profitable course of action. This scenario presents difficulties as to the standard of proof and risks unduly stretching competition law concepts. First, Article 101 TFEU does not prevent companies from using information available in the market to “adapt to existing and anticipated conduct of their competitors.”¹² Companies, foreseeing their rivals conduct, are free to change their prices.¹³ Second, it is only explicit collusion that is illegal under EU rules. An undertaking will only be held liable for breaching competition law when it “cooperate[s] with … competitors, in any way whatsoever, in order to determine a coordinated course of action relating to a price increase and to ensure its success […].”¹⁴

Conclusion

Companies should be aware that traditional cartel principles may be easily applied when algorithms can be shown to amount to mere cartel monitoring tools agreed by competitors. However, new theories of harm should be closely monitored by European Courts. For example, claiming algorithms are a means of tacit collusion and a breach of competition law in the absence of any awareness, intention, collusion or implementation mechanisms by competitors will be difficult to reconcile with basic principles of competition law, such as personal liability and standard of proof.

Antitrust Team