Algorithm cases


The world, as we know it, is rapidly digitising through breakthrough advancements in artificial intelligence (AI). Data sciences and AI are indeed the new reality and no longer a work of fiction as perceived in sci-fi movies. The increasing prevalence of data is driving businesses to develop and use AI to remain competitive in the market. Recent developments in machine learning have in fact elevated algorithms to a new level, enabling machines to analyse, act and learn with a human-like level of intelligence.

This characteristic of automated systems has drawn interest from global antitrust regulators as it enables businesses to achieve collusive outcomes without requiring an agreement in the traditional antitrust sense. Thus, the biggest challenge before regulators is detecting such price-fixing and tracing automated manipulation in the market, which is difficult, if not nearly impossible.

Further, AI can lead to unintended or autonomous collusion. Another associated challenge is establishing a causal link between algorithms and harm. Given these challenges, the overwhelming attention to algorithms globally seems justified. The Competition Commission of India (CCI) has also encountered issues related to algorithms in the airline and ride-hailing sectors. This article provides an overview of the anticompetitive usage of algorithms in India, and the extent to which such cases correspond to the global trend of such arrangements.

Algorithm cases

In 2018's Ride-hailing case,(1) the informant alleged that drivers have delegated their pricing decisions to common cab aggregator intermediaries (Ola and Uber) in a hub-and-spoke style cartel. The drivers, though independent contractors as submitted by Ola and Uber, followed the prices set by the pricing algorithms of the cab aggregators, and thus allegedly engaged in price-fixing to set supra-competitive prices.

Disagreeing, the CCI reasoned that cab aggregators are not associations of drivers and though the drivers charged the algorithm-determined prices, they did not agree among themselves to delegate the pricing power to the apps. Thus, they had no opportunity to coordinate fares while accepting rides. The CCI was also convinced that fares were dynamically priced, factoring in several parameters, such as:

  • distance;
  • traffic;
  • number of riders; and
  • drivers.

Such algorithmic pricing could be beneficial as well, as was noted in Webtaxi.(2) In this case, taxis belonging to several companies used Webtaxi, a booking platform in Luxembourg, which set taxi fares for them using price algorithms, based on several factors. The Luxembourg Competition Authority noted the arrangement as by-object restriction but found it beneficial, considering the algorithm-determined fares, based on digressive price per kilometre, would always be equal to, or lower than, the meter price and lower than those of competitors.

Similarly, in the two Airlines cases,(3) the CCI twice investigated the pricing of the leading airlines in India in 2014 and 2016 to examine if airfares were manipulated by algorithms. In both cases, the CCI enquired if common software was used or implemented based on a common understanding or such software resulted in price collusion in any way.

Much like the Ride-hailing case, the CCI found no evidence of collusion. The finding was supported by various considerations such as, use of different pricing software, and where the same software was used it was customised by historical inputs of the respective airlines. Further, the following factors drove the CCI to exonerate the airlines for the following reasons:

  • absence of price parallelism;
  • intervention from respective route analysts of airlines to decide airfares; and
  • evidence suggesting that pricing was influenced by the extant demand-supply conditions.

In one case, the CCI also noted the heavily fluctuating market shares of airlines during the alleged cartel period.

While it is too early to use these three cases to predict the future trajectory of algorithms and antitrust in India, they do shed light on the challenges associated with automated pricing cartels. They also echo the global trend – namely, most algorithm-related cases that have been found unlawful have one key thing in common: algorithms were used to enforce traditional anticompetitive arrangements. The cases of Trod (in the United Kingdom and the United States),(4) Casio (in the United Kingdom)(5) and Cigarette Cartel (in Spain),(6) among others, serve as examples of that commonality.

In the Trod case, Trod and GB Eye were penalised for agreeing not to undercut each other's posters and frames sold on Amazon UK by using repricing software. The case of Casio involved imposing resale price maintenance by Casio on retailers that sold its musical instruments online by using automated price-monitoring software. In Cigarette Cartel, a few leading tobacco companies and a distributor were fined in Spain and other jurisdictions for sharing real-time sales data and future pricing using software. All these cases indicate the application of textbook competition law principles to online distribution or service as in any traditional form of distribution.

Another concern about algorithms, outside of facilitating "explicit" or "traditional" collusion, is anticompetitive information exchange. The Eturas case(7) is a fitting example here, where Eturas, the owner of the online travel booking platform, sent a system-generated message to travel agents to cap the discounts. While Eturas was held liable for acting as a facilitator, Lithuania's highest court extended liability to only those travel agents who were aware of the restriction and did not oppose it.

It is interesting to distinguish the Ride-hailing case from the Webtaxi and Eturas cases. In the latter two cases, the lack of human intervention did not preclude the authorities from recognising the possibility of the facilitation of collusion without a meeting of minds or direct contact. In contrast, the CCI did not even acknowledge the existence of an agreement between cab aggregators and drivers, or drivers inter se in the Ride-hailing case, even though the drivers agreed with the cab aggregators to set prices, knowing well that other drivers also have similar arrangements with those aggregators. Thus, this could arguably have been viewed as a hub-and-spoke arrangement, given that drivers (competitors) had delegated price coordination and setting to the cab aggregators (hubs).

However, in the Airlines cases, the CCI was mindful of the absence of human interaction and the consequent trouble in establishing an agreement. Therefore, it emphasised in one of the Airlines cases that agreement could be deduced from several coincidences and indicia, which taken together and in the absence of an alternative explanation, could evidence an agreement.


The challenges illustrated above are merely the tip of the iceberg, as the prerequisite of establishing an agreement is especially tested in cases of self-learning AIs, that can result in autonomous tacit collusion. The key challenge remains as to whether such AI is just a tool used by humans or whether its ability to behave autonomously implies that its action cannot be attributable to a human operator. The latter could lead to a regulatory gap unless legal systems address that in advance.

As for future cases, given the opacity of AI systems, regulators globally are suggesting pre-emptive remedies, such as due diligence of algorithmic models of companies, so that companies disclose the workings of their algorithms and are held accountable for decisions made by those algorithms. In a similar vein, the Organisation for Economic Co-operation and Development (OECD)(8) also suggests a few approaches, such as price regulation, policies to make tacit collusion unstable and rules on algorithm design, while being mindful of the multi-dimensional nature of AIs which interface with intellectual property laws, consumer protection laws and data protection laws.

Despite the above concerns, it would be unwise to treat algorithms with only scepticism and hostility. If they can facilitate and automate collusion and unfairly discriminate against vulnerable consumers, they also enhance market efficiencies and allow firms to adapt intelligibly to market conditions in the following ways:

  • responding to stock availability;
  • capacity constraints;
  • reducing search costs and information asymmetries; and
  • providing price comparisons.

Therefore, although algorithms demand monitoring, especially in concentrated and homogenous markets that are already susceptible to collusion, the measures to address those concerns should be carefully crafted. This has been fairly displayed by the CCI, by not expressing any real concerns in the Ride-hailing and Airlines cases. However, it would be helpful for the CCI and other stakeholders to better appreciate the workings of AI by undertaking market studies and coordinating with experts in data sciences. A deeper understanding of the mysterious world of AI would facilitate a well-thought-out and balanced regulatory intervention.

As for companies using algorithms, "compliance by design", which can be achieved by documenting the purpose, inputs and technical workings of AIs, is the best option. The software developers are equally encouraged to customise algorithms to the individual customer's needs to avoid collusive outcomes and have robust confidential agreements to avert leakages of classified information.

For further information on this topic please contact Anisha Chand or Swati Bala at Khaitan & Co by telephone (+91 11 4151 5454) or email ([email protected] or [email protected]). The Khaitan & Co website can be accessed at


(1) Samir Agrawal v Competition Commission of India & Ors, Case No. 37 of 2018, decided on 6 November 2018. The CCI decision was upheld by the National Company Law Appellate Tribunal in Competition Appeal (AT) 11/2019 decided on 29 May 2020.

(2) Luxembourg Competition Authority decision 2018-FO-01, 7 June 2018.

(3) Alleged Cartelization in the Airlines Industry, Suo Motu Case No. 03 of 2015, decided on 22 February 2021 and Ms Shikha Roy v Jet Airways (India) Limited & Ors, Case No. 32 of 2016, decided on 3 June 2021.

(4) Online sales of posters and frames (Case 50223), the Competition and Markets Authority (CMA) decision of 12 August 2016. The CMA investigation followed similar investigations by the US Department of Justice in US v Daniel William Aston and Trod Limited (2016) and US v David Topkins (2015).

(5) Online resale price maintenance in the digital piano and digital keyboard sector (Case 50565-2) CMA decision of 1 August 2019.

(6) In 2019, Spain's National Commission of Markets and Competition fined the following companies a combined €57.1 million (S/DC/0607/17):

  • Philip Morris Spain;
  • JT International Iberia;
  • Altadis; and
  • Logista Integral Distribution Company.

(7) Case No. A-97-858/2016, Supreme Administrative Court of Lithuania decision of 2 May 2016.

(8) OECD (2017), Algorithms and Collusion: Competition Policy in the Digital Age.