Algorithmic Pricing and Competition: When Machine Learning Crosses Into Cartel Territory
Introduction
Price-fixing cartels have long been the most condemned form of anticompetitive behaviour. Their mechanism is straightforward: competitors agree to charge the same price, eliminate the competitive pressure that would otherwise drive prices down, and extract monopoly rents from consumers. Antitrust law in every major jurisdiction treats such agreements as per se illegal, requiring no analysis of market effect once agreement is established.
Algorithmic pricing, in its mature forms, achieves a functionally similar outcome without the meeting, without the agreement, and arguably without any conscious coordination at all. When competitors in a market each deploy dynamic pricing algorithms that observe the same market signals, including each other’s prices, and adjust their own prices accordingly in real time, the result can be sustained price alignment that is economically indistinguishable from cartel behaviour, even though no human from either company ever communicated their pricing intentions to the other.
In the online hotel accommodation, ride-hailing, and airline sectors, algorithm-driven price synchronisation has been documented by competition authorities in the United States, European Union, and, most recently, by the Competition Commission of India in a 2024 market study. The legal question is whether algorithmically achieved price coordination constitutes a prohibited agreement, a concerted practice, or a legal form of parallel behaviour, and what remedies are available in each case.
Legal Framework
The Competition Act, 2002, prohibits agreements between enterprises that have an appreciable adverse effect on competition under Section 3. An agreement is broadly defined in Section 2(b) to include any arrangement or understanding, whether or not formal or in writing. Section 3(3) treats horizontal agreements on price-fixing as presumptively having an appreciable adverse effect on competition, shifting the burden to the parties to rebut the presumption.
The prohibition on concerted practices, as developed in European competition law under Article 101 TFEU, has no precise equivalent in the Competition Act. The CCI has historically required some identifiable communication between parties. The algorithmic coordination case challenges this requirement: where coordination is achieved through independent algorithmic responses to shared market signals, there may be no communication to identify.
Judicial and Regulatory Developments
The Competition Commission of India’s 2024 market study on algorithmic pricing in the online travel accommodation sector concluded that algorithmic price synchronisation among major platforms was creating conditions economically equivalent to coordinated pricing. The study found that multiple accommodation aggregators were using pricing algorithms that shared a common dataset vendor and that the algorithms’ observable behaviour included price adjustments that tracked competitors’ prices within minutes of changes, a pattern inconsistent with independent pricing.
In the United States, the DOJ’s prosecution of RealPage, a software company providing algorithmic rent-setting tools to landlords, resulted in indictments in 2024 alleging that the algorithm constituted a mechanism for horizontal price coordination among competing landlords. The indictment theory is that by feeding proprietary rental data from each landlord into a shared algorithm that then recommended rents to all, RealPage facilitated an agreement to fix rents even without direct communication between landlords.
Contemporary Issues and Analysis
Three distinct algorithmic coordination scenarios present different legal challenges. In the first, competing firms independently program their algorithms to respond to competitors’ prices, producing conscious parallelism through purely independent action. Under traditional antitrust doctrine, independent parallel pricing is not prohibited in the absence of an agreement. This scenario creates no clean antitrust violation, but it produces the same consumer harm as a cartel.
In the second scenario, competing firms each use a pricing algorithm provided by a common third-party vendor, whose algorithm is designed in a way that produces coordinated outputs across all subscribers. Here, the antitrust theory is that the vendor is a horizontal facilitator, and each firm’s subscription constitutes participation in an agreement to coordinate pricing through the common mechanism.
In the third scenario, competing firms deploy AI-based pricing algorithms that, without any common vendor or shared mechanism, learn through reinforcement to coordinate prices with competitors because coordination is the strategy that maximises individual profits. Economic modelling has demonstrated that in oligopolistic markets, reinforcement learning algorithms trained to maximise revenue will independently discover and sustain supracompetitive pricing as an equilibrium strategy. This autonomous discovery of coordination represents the most challenging case for antitrust theory.
Comparative and International Perspective
Germany’s Bundeskartellamt produced a 2024 working paper proposing that competition law should be interpreted to prohibit algorithmic coordination where: the algorithm is designed to monitor competitor prices and respond to maintain alignment; the coordination produces supracompetitive prices; and a reasonably sophisticated analysis of the market structure confirms that the alignment cannot be explained by independent rational responses to common cost or demand signals.
Practical and Policy Implications
For Indian businesses using dynamic pricing algorithms, the CCI’s 2024 market study signals that regulatory attention is focused and enforcement may follow. Companies in accommodation, ride-hailing, food delivery, and e-commerce should conduct internal algorithmic audits to assess whether their pricing behaviour exhibits characteristics of coordination and whether their algorithm vendor relationships create facilitating mechanism exposure.
Suggestions and Reforms
The Competition Act should be amended to address algorithmic coordination explicitly. A new explanation to Section 3 should provide that a concerted practice includes coordination achieved through common algorithmic mechanisms, shared data vendors, or mutual awareness of algorithmic responses among competitors, whether or not explicit communication occurred.
The CCI should develop Algorithmic Pricing Guidelines modelled on the OECD’s 2024 Roundtable recommendations, specifying the conduct that raises competition concerns, the analytical framework for investigating algorithmic coordination, and the standards by which companies can demonstrate that their pricing algorithms are consistent with independent competitive behaviour.
Conclusion
Competition law’s prohibition on price-fixing was designed to protect the market process, the dynamic through which independent competitors discover prices that reflect actual costs and consumer valuations. When algorithms eliminate that process by achieving coordination more efficiently than any human cartel could, they do precisely what the law was meant to prevent, regardless of whether a telephone conversation took place. The law must evolve to match the technology, not because algorithms deserve stricter regulation than humans, but because the harm they can cause is real, widespread, and currently underenforced.
