Businesses no longer need to agree on prices themselves. Instead, they are using self-learning algorithms. Where does unlawful coordination begin? / by Markus Schöner and Denis Schlimpert
Companies are increasingly using algorithms to determine prices. Especially in online shops, Artificial Intelligence (AI) can play a significant role through the use of self-learning pricing algorithms – and the topic has been in the crosshairs of competition authorities for some time (F.A.Z. Einspruch of 14 March) (http://einspruch.faz.net/einspruch-magazin/201803-14/auf-der-suche-nach-dem-besten-preis/63369.html).
Competition law prohibits competitors from making agreements on their prices . They are also not allowed to exchange information to coordinate their pricing behaviour. However, companies are permitted to monitor competitors’ public behaviour and make intelligent adjustments to their prices. This is lawful parallel behaviour.
Self-learning algorithms are on the interface between unlawful coordination and lawful parallel behaviour. They are computer programmes that record and process information, and are continuously learning. Self-learning pricing algorithms can learn, without any human interaction, that it may be financially beneficial to match the published online pricing behaviour of other market participants. The result would be adjusted prices without the need for any agreement or coordinated behaviour. The decision to use an algorithm is a unilateral measure not covered by the prohibition of cartels.
At the same time, the EU Competition Commissioner, Margrethe Vestager, has declared publicly on several occasions that companies cannot hide behind computer software to create cartels. In fact, companies are also responsible for actions carried out by automated systems. Pricing algorithms should be programmed in such a way as to preclude collusion. The president of the German Federal Cartel Office, Andreas Mundt, has also warned companies: “Algorithms are not made by the dear God.”
Nonetheless, in its latest report of July 2018, the Monopoly Commission, an independent committee of experts that advises the German federal government on competition matters, has warned against hasty changes to the law. It argues instead for stronger monitoring of the situation using sector inquiries by the competition authorities. Ideally, consumer protection authorities should have the right to initiate such inquiries.
The Monopoly Commission is thus following the Organisation for Economic Co-operation and Development, which had initially only proposed market studies in a report dating from 2017. Amending competition law should only be considered with progressing research on algorithms. In view of the complexity of algorithms and the currently low use of AI-based algorithms, such a cautious approach seems reasonable.
However, even today, competition law is in no way powerless with regard to algorithms: Companies that wish to use (pricing) algorithms must ensure that this is done in a manner consistent with competition law.
Of course, companies should not make arrangements with other companies to use certain algorithms, which then in turn implement coordinated pricing behaviour. Accordingly, companies should not agree to engage the same IT service provider, which then supplies the same algorithm.
A company also should not use an algorithm to transmit its confidential information to competitors or to process confidential information from competitors. Ultimately, this would be coordinated behaviour based on a prohibited exchange of information.
Caution is also required if the aim is to monitor a customer’s resale prices using algorithms, for example in the relationship between a brand manufacturer and the online retailer it supplies. Merely monitoring does not result in an agreement or coordination. However, the data generated in this way must not be used to influence the pricing behaviour of customers. That would be the case, for example, if the manufacturer were to use the pricing it has monitored, which, in its opinion, is too low, to put pressure on its customers.
As a result, companies are well advised to not just wait and see how AI-based algorithms and the ongoing debate about them develop in the future. They should first ensure that the algorithms they use are compliant with the law.
Dr. Markus Schöner is a partner and head of the Competition Law Department at CMS in Germany. Dr. Denis Schlimpert is a lawyer at the firm’s Hamburg office.
First published in F.A.Z. Einspruch, 25 July 2018 http://einspruch.faz.net/einspruch -magazin/2018-07-25/wer-bestimmt-den-preis/119067.html(not updated with developments and/or reports after the first publication)