Coded Collusion: Algorithmic Pricing In Indian Aviation And Structural Failure Of Competition Law

This blog addresses three connected problems. Firstly, it asks why Section 3 of the Competition Act, 2002 as currently drafted cannot reach algorithmic pricing even when market outcomes look indistinguishable from cartel behavior. Secondly, it evaluates and examines CCIs own precedents, particularly its jurisprudence in Shikha Roy v. Jet Airways, tell us about the evidentiary wall that algorithmic coordination hides behind. Thirdly, it assesses whether India's existing regulatory architecture spread across the CCI, the DGCA, and the Ministry of Civil Aviation is institutionally capable of identifying and addressing the kind of invisible cartel that algorithmic pricing produces, or whether that task requires structural reform.

Part II explains how reinforcement learning pricing engines operate as active market shaping instruments and why they produce supra-competitive outcomes without any explicit coordination. Part III analyses the Section 3 evidentiary framework and the agreement centric gap that algorithmic convergence exploits, using Shikha Roy as the anchoring precedent. Part IV maps India's fragmented three-regulator architecture against the EUs emerging algorithmic governance model under the Digital Markets Act and the EU AI Act. Part V draws three concrete reform recommendations targeted at the CCI, Parliament, and the aviation sector specifically.

From Revenue Tools to Tacit Coordinators: How AI Pricing Engines Silently Restructure Market Competition

The contemporary airline pricing systems are not considered as passive revenue mechanisms rather they operate as active intelligence systems. These platforms usually use reinforcement learning, which is considered as a technique in which algorithm iteratively tests pricing strategies and then optimizes towards the most profitable outcome by analyzing how competitors respond. Calvano's landmark 2020 experimental study demonstrated that algorithms independently operating in oligopolistic markets independently converge to supra competitive prices, not through coordination but thorough parallel computational learning in which each algorithm inferring the counter parts strategy and adapting accordingly. The aviation industry in India is structurally focused and boosted. Few carriers compete over high traffic routes like Delhi-Chandigarh or Delhi-Mumbai, each one of them treating algorithmic pricing data as an input signal. This risk produces a computed echo chamber where each generated algorithmic output becomes the algorithmic input, which generates the price alignment without any underlying agreement. This echo chamber premise, must be construed contrary to the findings of the Director General in Shikha Roy itself, where the DG, after carefully analyzing real route level pricing data across the very corridors invoked to illustrate algorithmic convergence, found no evidence of coordinated conduct, concluding instead that the observed price movements were in line with the independent competitive responses to shared market conditions. The same was acknowledged in CCIs Market Study in 2025, which states that the tacit collusion can be facilitated by algorithmic tools which allows firms to track and react to competitor's behavior in real time.

The Agreement Nobody Made: Section 3, the CCI's Evidentiary Threshold, and the Algorithmic Loophole

Under Section 3(1) of the CCI Act, 2002 agreements between the enterprises are prohibited which causes or likely to cause an appreciable adverse effect on competition. Section 3(3) also creates a presumptive, near per se prohibition for horizontal arrangements which involves price fixing, market allocation or bid rigging. Section 2(b) defines the term 'agreement' which presents a critical evidentiary threshold by including any arrangement or understanding, whether or not formal, informal, written or oral. Algorithmic pricing structurally evades this threshold i.e. “requirement of an arrangement or understanding”, because two competing pricing engines arriving at identical fares by independently processing the same publicly available market signals share no mutual intent, no private communication, and no conscious expectation of each other's conduct, they converge not by agreeing, but by computing.

Furthermore, CCIs order in Shikha Roy v. Jet Airways (India) Limited and Others is the most instructive and paramount Indian precedent on this question. It was alleged by the informant that five domestic carriers like Jet Airways, SpiceJet, IndiGo, GoAir, and Air India had cartelized on the routes like Delhi-Chandigarh and Delhi-Amritsar by simultaneously raising fares. CCI after carefully analyzing the pricing data, concluded that although sufficient parallel movements of the prices were observed, they are not adequate enough to be considered within the definition of agreement as defined under Competition Act. The Commission opined that parallel conduct alone does not constitute substantial evidence strong enough to support an inference of coordination in the absence of any additional factor such as explicit communication, information exchange or any structural interdependence beyond mere oligopolistic reaction.

CCI, as in the other contexts, has accepted indirect evidence of coordination where plus factors such as trade association communication or documented information exist, as was seen in the Cement Cartel proceedings of 2012. In the context of algorithmic use, no such plus factor exists by design, as they do not communicate but converge. Under the current Section 3 of the Competition Act as made clear by Shikha Roy, convergence without communication is legally permissible.

The deeper structural lacuna is Section 3's agreement centric framework which is designed specifically for markets where anti-competitive coordination is produced by human intent. Algorithmic coordination usually produces cartel like outcomes as a systemic byproduct rather than a deliberate objective coordination as it operates on a common data emerging from competitive system which makes it categorically different. This legislative gap is not yet addressed by any statutory regulation or CCIs enforcement practice.

Three Regulators, No Answer: India's Fragmented Framework and the EU's Algorithmic Blueprint

Under the Indian context, Directorate General of Civil Aviation is vested with the authority under Aircraft Act, 1934 and Aircraft Rules, 1937 to supervise safety and airworthiness with fare regulation falling expressly outside its statutory mandate. National Civil Aviation Policy, 2016, assumes that competition will self-regulate fares without being prescriptive in nature. The European Union's regulatory framework offers a more systematic and structurally grounded regulatory model. EU's Digital Market Act (DMA), which came into force on March 2024, imposes mandatory obligations on designated 'gatekeepers' under Articles 5 and 6. Under Article 6(2) prohibition is placed on the use of non-public business data which is generated by commercial users against others. The EU AI Act, effective from August 2024, on the other side explicitly categorizes certain algorithmic systems which affects the pricing in commercial context as 'high risk' under Annex III, mandating pre deployment risk management systems under Article 9. The analytical focus is thus shifted from pricing outcomes to the architecture of the algorithmic process itself. India's approach towards the algorithmic pricing inverts this logic, and EU's framework highlights that this inversion is a regulatory choice rather than a structural inevitability.

Modernise or Concede: Reforming India's Competition Law Before the Algorithmic Cartel Becomes Permanent

The problem of algorithmic pricing in India's aviation sector is not at is core, a technological problem, it's a structural problem of legal cartelization. Section 3 of the Competition Act, 2002 was crafted specifically to cater to the markets where anti-competitive coordination is a product of human intent. Algorithmic coordination is an emergent, systemic outcome of rational computational behaviour operating within a legally permissive framework. The existing legal regulation cannot reach it, and this gap is fulfilled by Shikha Roy precedent with rare precision. A Section 49 investigation must go one level deeper into the objective functions, training data, and competitor-signal inputs of each airline's pricing engine because that is where coordination, if it exists, will be found. The algorithmic pricing presents a structurally identical but computationally more opaque evidentiary problem if CCI declines to infer a cartel even from parallel pricing on specific routes by the same competing carries as no additional plus factor establishing conscious parallelism was available.

Three analytical conclusions follow from this diagnosis. Firstly, CCIs Market Study on AI and Competition it creates foundation for sectoral intervention under its Section 49 market study regulation. Secondly, Section 3 (3) governs horizontal arrangements and operates through a near per se standard which means once a qualifying arrangement is established, anti-competitive effect is presumed rather than proved, placing a significantly higher and faster evidentiary burden on the parties than the full rule of reason analysis required under Section 3(4) for vertical agreements. Algorithmic cartelization is considered as a horizontal conduct and must remain a matter dealt under Section 3(3). Parliament should introduce a deeming provision within Section 3(3) specifically not Section 3(4) treating sustained, unexplained price alignment produced by algorithms operating on a shared or mirrored data inputs as raising a rebuttable presumption of coordination, with the burden of disproving that presumption shifting to the airlines. Locating this deeming provision within Section 3(3) is not a drafting technicality it is a substantive choice that preserves the per se character of cartel prohibition and prevents algorithmic coordination from being diluted into a slower, more uncertain rule of reason enquiry that was designed for an entirely different category of commercial conduct. Thirdly, The EU AI Act's pre-deployment conformity assessment model offers India a technically feasible and proportionate compliance architecture but transposing it into the Indian framework requires engaging with three specific institutional questions as to which authority administers it, under which existing statutory power, and through what procedural mechanism. Now this framework can be incorporated into Indian framework mainly through three ways. Firstly, on statutory authority the CCI already possesses the power to frame regulations governing market conduct under Section 64 of the Competition Act read with Section 49, which empowers the Commission to issue guidelines on competition advocacy and market studies. A pre-deployment disclosure obligation for algorithmic pricing systems does not require fresh legislation it can be framed as a regulation under Section 64 requiring airlines above a defined market-share threshold to file a Conformity Assessment Report before deploying or materially updating any fare-optimisation algorithm. This is the precise mechanism through which the EU's national competition authorities have operationalised the DMA's algorithmic transparency obligations without requiring separate Parliamentary action.

Secondly, on the content of the Conformity Assessment Report, drawing directly from Article 9 of the EU AI Act's risk-management framework, the Indian CAR should require disclosure of three things the data inputs feeding the algorithm, specifically whether they include real-time competitor fare signals, the learning objective of the system, including whether profit maximisation is conditioned on competitor behaviour; and an independent audit certificate confirming the system cannot access a competitor's non-public pricing data. These three disclosures replicate the architectural transparency that the EU framework mandates, translated into the specific data-access concerns that the Shikha Roy investigation lacked.

Thirdly, on the triggering threshold, unlike the EU's gatekeeper designation under the DMA which applies to platforms of a defined scale India's aviation market is concentrated enough that a simpler route level threshold is more appropriate. Any airline holding fifteen per cent or more of seat capacity on any single domestic route should be subject to the CAR obligation on that route. This is a proportionate and administrable threshold that directly targets the concentrated trunk routes Delhi–Mumbai, Delhi–Chandigarh where algorithmic convergence risk is empirically highest, without imposing compliance costs on genuinely marginal carriers. The Indian incorporation of the EU model therefore does not require legislative amendment, a new regulatory body, or a wholesale adoption of the EU's gatekeeper architecture it requires the CCI to exercise statutory powers it already holds, directed at a disclosure obligation that is architecturally specific, route level in its trigger, and independently audited in its verification.

Airlines that are big enough (crosses a “market threshold”) should be forced to upfront that their computer pricing systems are not mainly built to spy on competitor's price and automatically react to them.

The algorithmic cartel is not invisible because it is absent. It is invisible because the legal tools meant to regulate it were designed for a different era. A framework that assumes human intent, demands explicit and clear agreement, and looks at published prices rather than computational logic will repeatedly fall short against the next generation of anti-competitive behaviour. India's competition law must now choose whether to modernize its approach or quietly bear the consumer welfare costs of that failure. The longer this decision is delayed, the more deeply entrenched the algorithmic advantage becomes, and the greater the ultimate cost of inaction.