Written by Nupur Bajoria

ABSTRACT

Algorithmic pricing exposes a structural weakness in India’s cartel framework: coordinated, supra-competitive pricing may arise without human communication. This paper argues that the Competition Commission of India’s (CCI) treatment of algorithmic pricing in Ms. Shikha Roy v. Jet Airways and In Re: Alleged Cartelization in the Airlines Industry was too dependent on the absence of express or tacit agreement. Read through the plus-factors approach in Express Industry Council of India v. Jet Airways, the relevant question is whether pricing is better explained by independent competition or by a computational environment capable of producing coordination. Shared pricing software, repeated parallel movements, reliance on historical competitor behaviour, and the possibility that one airline’s pricing signal becomes another’s algorithmic input are not individually conclusive; cumulatively, they justify deeper scrutiny. The CCI’s 2025 Market Study on Artificial Intelligence and Competition provides the institutional basis for that inquiry. The paper therefore proposes a plus-factors test in which objective functions form one category of circumstantial evidence that may, when considered cumulatively with other factors, support an inference of tacit collusion.

I. Introduction: From “agreement” to computational coordination

As stated in the market study, pricing algorithms can monitor rivals, ingest historical data, react to competitor signals and, in self-learning systems, optimise against market responses. Two firms c therefore reach the same supra-competitive price without communicating; the causal mechanism can be computational rather than communicative.

India’s Competition Act, 2002 nevertheless begins from an agreement-centric architecture. Section 3(1) prohibits agreements causing or likely to cause an appreciable adverse effect on competition (“AAEC”), while Section 2(b) defines “agreement” to include an arrangement, understanding or concerted action. Section 3(3) presumptively treats horizontal price-fixing agreements as anti-competitive. The Competition (Amendment) Act, 2023 also brought hub-and-spoke arrangements within the presumptive framework where a non-competitor participates or intends to participate in furtherance of the agreement. These provisions strengthen the statute, but do not resolve the harder case where coordination is generated by competing algorithms without a legally cognisable human instruction. The CCI’s chairperson, Ravneet Kaur, has recently acknowledged this issue, stating that “Companies are using AI to coordinate prices without explicit agreements, and are engaging in algorithmic discrimination under the garb of dynamic pricing.”

The CCI’s Market Study on Artificial Intelligence and Competition, released on 6 October 2025, now identifies algorithmic collusion, price discrimination, predatory pricing and opacity as emerging risks. It also recommends competition-oriented self-audits, documentation, monitoring and technical capacity. However, the Study does not take the next evidentiary step: the CCI should make the plus-factors test the starting point for deciding whether algorithmic pricing is independently competitive or functionally coordinated.

II. The two airline matters: what the CCI found and what it did not test

In Ms. Shikha Roy v. Jet Airways, the DG examined unusually high fares during the February 2016 Jat protests. The CCI recognised that automated pricing could generate parallel price movements without human contact, but closed the matter because parallel behaviour, without clearer evidence of coordination, did not establish a Section 3(1) violation. The Director General (DG) also found that airlines used different, customised algorithms and that route analysts retained the final decision over inventory allocation.

In Re: Alleged Cartelization in the Airlines Industry, the CCI examined third-party revenue-management software and dynamic pricing on major domestic routes. It found no conclusive cartel because route analysts made final pricing decisions and market shares varied. However, it noted that individual airlines used multiple algorithms and some of them overlapped across airlines, including Navitaire. Even with manual intervention in presenting final ticket fares, the bench did not consider the impact of algorithmic recommendation itself incorporating historical rival behaviour and thereby creating an echo chamber of influencing airlines’ pricing.

Moreover, the weakness is not that the CCI ignored algorithms, but that the absence of explicit coordination was treated as substantially dispositive without testing whether the algorithmic environment supplied a functional substitute for communication. A route analyst exercising discretion over a recommendation is not necessarily independent if the recommendation incorporates historical rival prices. In other words, if two firms use different algorithms, but those algorithms systematically ingest rival prices and react to one another, when can that conduct cease being ordinary competitive adaptation and become evidence of concerted action? Furthermore, this concerted action can be done even with publicly available data, with relatively simple algorithmic models that do not require proof of the algorithms to exchange private data amongst them because, over time, the algorithms learn that undercutting competitors triggers price wars that hurt everyone’s margins.

This also distinguishes the airline cases from Samir Agrawal v. ANI Technologies. There, the CCI, NCLAT and Supreme Court found no prima facie contravention because the alleged spokes, the drivers, had no opportunity to coordinate with each other, and the algorithm responded to demand and supply. That conclusion does not establish that algorithmic pricing is necessarily competitive whenever firms lack express communication. The airline cases present a different question: whether competing pricing systems repeatedly process common or rival-derived information so that collusion becomes a better explanation than independent competition.

III. The plus-factors test supplies the missing bridge

In Express Industry Council of India v. Jet Airways, the Commission rejected the proposition that parallel fuel-surcharge movements could be dismissed as ordinary oligopolistic conduct. It relied cumulatively on repeated identical revisions in amount and timing despite different commercial considerations, absence of documented methodology or cost analysis, and inconsistent explanations. These circumstances operated as “plus factors” because they made independent decision-making less plausible.

The Supreme Court’s reasoning in Rajasthan Cylinders and Containers Ltd. v. Union of India and the appellate analysis reflected in Excel Crop Care reinforce the same proposition: parallelism is not itself proof of collusion, but its evidentiary value depends on whether the conduct departs from what normal market conditions would rationally produce.

Algorithmic pricing should be assessed through that same logic. The inquiry should ask, as adopted with the help of Calvano et al.: (i) are prices often parallel in magnitude, timing or direction beyond what common demand shocks can explain; (ii) do algorithms ingest competitor prices, historical rival behaviour or common third party data; (iii) do competing firms use same or materially overlapping pricing software; (iv) do firms have no contemporaneous cost, demand or route-specific explanations; (v) do algorithmic outputs consistently track or reinforce rival prices; (vi) is human override genuinely independent; and (vii) do audit logs, objective functions, historical data and system constraints reveal incentives capable of producing supra-competitive convergence?

While the above plus factors are suggested keeping in mind algorithmic pricing in the airline industry, the Competition Bureau in Canada has acknowledged that algorithmic pricing could also in the future impact affordability of daily essentials in sectors such as groceries, transportation and housing. Therefore, no single factor should establish liability. The point is cumulative; where several factors cohere, the question should shift from “show us the communication” to “show us why independent competition is the better explanation.” This is not a presumption that every parallel algorithm is a cartel.

IV. Why the 2025 CCI AI Market Study should change enforcement practice

The 2025 Market Study is significant because it moves the CCI toward the architecture of the pricing system. It distinguishes monitoring, parallel/hub-and-spoke, signalling and self-learning algorithms, and recognises that self-learning systems can converge on collusive outcomes without explicit human intervention. Its compliance checklist also asks whether algorithms use proprietary data, whether algorithmic decisions are logged, whether human oversight exists for significant price changes, and whether competition-risk assessments and historical records are maintained.

That framework should guide both Section 49 market studies and, where warranted, subsequent Section 26 investigations. In future algorithmic pricing cases, the CCI should reconsider the algorithmic decision-making chain. It should assess from the objective function and data inputs, through competitor signals and algorithmic recommendations, to human intervention and the final prices. The relevant inquiry should therefore extend beyond the observed pricing outcome to the computational process that generated it, since it is within that process that evidence of coordination may emerge.

The 2023 amendment is already conceptually helpful because Parliament recognised that modern coordination may operate through a hub which is not itself a competitor. The CCI’s AI study now recognises algorithmic coordination as a competition risk. The remaining step is to treat circumstantial algorithmic evidence as legally meaningful even where the computational pathway produces no traditional “meeting of minds.”

In the EU, the issue of pricing algorithms was considered by the Court of Justice of the EU in the E-TURAS case. E-TURAS, an online travel booking system, sent messages to its travel agents through the online system announcing technical restrictions to its pricing algorithm, capping discounts at 3%. The Court of Justice confirmed that even though travel agencies did not formally respond to the message, the fact that they were aware of such a message, did not distance themselves from it and have

subsequently continued to use the system, such agencies may be liable for the price-fixing cartel under Article 101(1) of the Treaty on the Functioning of the European Union (TFEU). We can see the plus factors test being used here in spirit, that there was a common AI tool, collusion was between all the enterprises even if there was no express agreement, and human override was not independent.

V. Conclusion: competition law must follow the mechanism of coordination

The central mistake in treating algorithmic pricing as an ordinary cartel problem is to make actual and tacit agreement the only criteria for coordination. In a conventional cartel, communication explains how firms aligned. In an algorithmic market, the relevant question is whether the pricing architecture makes alignment systematically likely and whether the resulting conduct is inconsistent with independent competition. The existing Section 3 evidentiary framework can accommodate algorithmic evidence, but the existing jurisprudence does not yet explain how algorithmic evidence should be used to infer the underlying agreement or concerted action where collusion emerges from automated interaction.

The two airline matters demonstrate the problem. The CCI found manual intervention, differentiated algorithms and variable market shares, but those observations did not resolve whether rival-derived historical information, overlapping software and repeated pricing patterns could generate functional coordination. Applied rigorously, the plus-factors approach from Express Industry Council provides the better evidentiary framework: parallelism remains insufficient by itself, but becomes probative when accompanied by technical and commercial circumstances making independent pricing implausible. The CCI should therefore institutionalise a plus-factors screen for algorithmic pricing. This preserves the requirement of proof while avoiding an artificial insistence on a human meeting of minds where coordination occurs through computation. The objective is not to prohibit dynamic pricing or algorithmic efficiency. It is to ensure that firms cannot obtain cartel-like outcomes merely because coordination is embedded in code rather than conversation. The 2025 AI Market Study supplies the technical foundation; the next step is to translate it into competition-law methodo