This piece of the article is authored By:- Gaurav Kumar Singh & Soumya Manish Bohra, student at The West Bengal National University of Juridical Sciences.
The emergence of artificial intelligence as a structural feature of commercial markets has introduced a form of anti-competitive conduct that defies the foundational logic upon which Indian competition law was constructed. The Competition Act, 2002 was designed to detect and penalise cartels formed by human beings — enterprises that consciously communicated, agreed, and coordinated their pricing and market behaviour in contravention of the law. The legislative imagination that produced Section 3 of the Act envisaged boardroom conspiracies, written price-fixing agreements, and market-sharing arrangements concluded by identifiable corporate actors exercising deliberate judgment. It did not and could not have anticipated a world in which competing enterprises deploy self-learning pricing algorithms that, without any human communication, instruction, or express agreement, independently arrive at coordinated above-competitive prices through a process of mutual algorithmic observation and machine-level adaptation. This phenomenon — variously described in competition law scholarship as algorithmic collusion, tacit algorithmic coordination, and machine-to-machine cartelisation — is no longer a theoretical abstraction. India’s ride-hailing markets, e-commerce platforms, digital advertising ecosystems, and financial services sector are already exhibiting pricing dynamics consistent with AI-driven coordination, and the Competition Commission of India’s own Market Study on Artificial Intelligence and Competition, released in October 2025, has expressly acknowledged that pricing algorithms, particularly self-learning and signalling algorithms, may produce coordinated pricing behaviour across markets without any legally cognisable agreement between the firms involved.
The urgency of this challenge is sharpened by the scale and trajectory of India’s AI market. According to the CCI’s market study, India’s AI sector has doubled from $3.2 billion in 2020 to $6.05 billion in 2024, and is projected to reach $31.94 billion by 2031. Against this backdrop, approximately 37% of surveyed startups participating in the CCI study expressed concern about AI-facilitated collusion, underscoring that algorithmic coordination is not a hypothetical future harm but a present market reality. The Competition (Amendment) Act, 2023 introduced the hub-and-spoke cartel framework through a new proviso to Section 3(3) of the Competition Act, representing the most significant substantive reform to India’s cartel enforcement architecture since the Act’s original enactment. This article argues that while the 2023 Amendment is a meaningful step forward, it is structurally insufficient to address the most dangerous and most prevalent form of AI-driven price coordination — namely, the scenario in which competing firms independently deploy a shared AI pricing engine, and the AI system itself, without any human facilitation, orchestrates above-competitive pricing outcomes that cause demonstrable harm to consumers and market competition. The article proceeds through five analytical heads: the doctrinal architecture of Section 3 and the agreement requirement; the 2023 Amendment’s hub-and-spoke framework and its application to AI; comparative jurisprudence from the European Court of Justice and the United States; the evidentiary void in detecting algorithmic collusion; and, finally, the prescriptive framework that India’s competition policy must urgently construct.
I. Section 3 of the Competition Act, 2002 and the Structural Problem of the Agreement Requirement
The foundational prohibition against anti-competitive agreements in Indian competition law is located in Section 3(1) of the Competition Act, 2002, which provides that no enterprise or association of enterprises or person or association of persons shall enter into any agreement in respect of production, supply, distribution, storage, acquisition, or control of goods or provision of services which causes or is likely to cause an appreciable adverse effect on competition within India. Section 3(3) then enumerates specific categories of horizontal agreements — those relating to price-fixing, market allocation, bid-rigging, and production control which are presumed to have an appreciable adverse effect on competition and are therefore treated as per se violations of the Act, subject only to the efficiency defences available under Section 19(3). The entire liability architecture of Section 3 rests upon the concept of an “agreement,” which is defined under Section 2(b) of the Act to include any arrangement, understanding, or action in concert, whether or not such arrangement, understanding, or action is formal or in writing, or is intended to be enforceable by legal proceedings.
The definition under Section 2(b) is deliberately broad and captures informal understandings and tacit arrangements, going considerably beyond the narrow contractual definition. However, even this expansive formulation requires, as a threshold matter, some form of communication, shared understanding, or concerted action between identifiable human or corporate actors. The problem with algorithmic collusion is precisely that the AI systems involved may reach coordinated pricing outcomes through a process of independent algorithmic learning, in which each competing firm’s algorithm observes publicly available market signals — including the pricing outputs of competitors’ algorithms — and adjusts its own prices accordingly, with neither firm ever communicating with the other, sharing data with the other, or even being aware of the coordination taking place at the level of the algorithm. This is the phenomenon that competition economists call a “messenger algorithm” or a “self-learning algorithm,” and it is structurally distinct from the human-mediated tacit collusion that Section 3 was designed to address. The CCI itself has acknowledged in its 2025 Market Study that the current legal framework treats AI-driven pricing parallelism as “conscious parallelism” or “tacit collusion,” which is generally not actionable under Indian law unless accompanied by additional “plus factors” — additional evidence of intent, communication, or concerted action that elevates mere parallel conduct to the level of a Section 3 violation. This is the evidentiary void at the heart of India’s algorithmic collusion problem: where an AI system produces cartel-like outcomes without any human agreement, the injury to competition is identical to that caused by a traditional cartel, but the legal remedy is absent.
II. The Hub-and-Spoke Framework Under the Competition (Amendment) Act, 2023: Reach and Limitations
The Competition (Amendment) Act, 2023, which received Presidential assent on 11 April 2023, introduced through a new proviso to Section 3(3) the explicit recognition of hub-and-spoke cartel arrangements in Indian competition law. Under the pre-amendment framework, Section 3(3) applied only to parties engaged in identical or similar trade horizontal competitors — leaving the role of intermediaries and platform facilitators in coordinating competitor conduct in a structural ambiguity. The 2023 Amendment addressed this gap by providing that where an enterprise, association of enterprises, person, or association of persons, even if not engaged in identical or similar business, participates or intends to participate in furtherance of a prohibited agreement under Section 3(3), such person shall be presumed to have caused an appreciable adverse effect on competition and shall be liable under the same provisions that govern the horizontal competitors constituting the “spokes” of the arrangement. The significance of this amendment for AI-driven collusion lies in its potential to capture the role of an AI software provider a technology company that supplies a shared pricing engine to multiple competing firms — as the “hub” in a hub-and-spoke cartel, with those competing firms constituting the “spokes.”
The landmark case of Samir Agrawal v. Competition Commission of India, decided by the National Company Law Appellate Tribunal in May 2020 and affirmed by the Supreme Court of India in December 2020, provides the most directly relevant Indian judicial treatment of algorithmic pricing to date. In that case, the appellant alleged that Ola and Uber had engaged in price-fixing by deploying common algorithmic pricing systems — specifically, dynamic surge pricing algorithms that resulted in coordinated pricing between competing cab aggregators and between the platforms and their driver-partners. The NCLAT dismissed the complaint, holding that the algorithmic pricing employed by Ola and Uber did not constitute cartelisation under Section 3 of the Act because there was no evidence of any agreement, understanding, or concerted action between the competing platforms or between the platforms and their drivers. The Supreme Court affirmed the NCLAT’s ruling, additionally holding that the appellant lacked the standing to bring a complaint before the CCI as an informant under Section 19(1)(a) without demonstrating a direct harm, thereby raising an additional procedural barrier to algorithmic collusion complaints brought by consumer-informants.
The Samir Agrawal decision is instructive as a marker of the pre-amendment doctrinal baseline: the absence of an express agreement between competing firms using similar algorithmic systems was, in itself, fatal to a Section 3 claim. The 2023 Amendment’s hub-and-spoke provisions carry the potential to alter this outcome prospectively, insofar as an AI software company whose pricing engine is deployed by competing firms and whose design choices effectively coordinate their pricing conduct could be characterised as the “hub” of a prohibited arrangement. However, the Amendment’s provisions do not resolve the deeper structural problem, because even under the expanded Section 3(3) framework, a hub-and-spoke cartel requires that the hub — the AI developer or software provider — participates or intends to participate in furtherance of the prohibited agreement. Where the AI pricing engine coordinates competitor pricing through emergent machine learning rather than through any deliberate design choice, proving that the software provider “participated” in or “intended to participate in furtherance of” a cartel is an evidentiary challenge of the highest order. The Amendment resolves the structural question of intermediary liability but leaves wholly unaddressed the deeper question of algorithmic intent — and it is in precisely that unaddressed space that the most dangerous forms of AI-driven collusion operate.
III. Comparative Jurisprudence: Eturas, RealPage, and the International Enforcement Landscape
India does not exist in a regulatory vacuum, and the doctrinal challenges posed by algorithmic collusion have been confronted — with varying degrees of success — by competition authorities and courts in the European Union and the United States. The jurisprudence from these jurisdictions offers both instructive precedent and cautionary lessons for the development of India’s enforcement framework.
The landmark case from the European Union is the judgment of the Court of Justice of the European Union in Case C-74/14, Eturas UAB and Others v. Lietuvos Respublikos konkurencijos taryba (Lithuanian Competition Council), decided on 21 January 2016. The case arose from the conduct of Eturas, the operator of a shared online travel booking platform used by multiple competing Lithuanian travel agencies. The platform operator introduced a technical restriction that capped the discount rates that travel agencies could offer to consumers, and sent a system message notifying agencies of this change. The CJEU held that where competing enterprises use a shared IT system and have knowledge of a communication or technical modification that produces anti-competitive coordination, they may be presumed to have participated in a concerted practice for the purposes of Article 101 of the Treaty on the Functioning of the European Union, unless they took active steps to distance themselves from the coordination and informed competition authorities. The CJEU’s crucial contribution in Eturas is the articulation of a knowledge-based presumption: actual explicit communication between competitors is not required for a concerted practice to exist where a shared platform creates the infrastructure for coordination and the participants have knowledge of its coordinating function. This Eturas presumption carries direct implications for the Indian context, because it suggests a framework under which the CCI could treat competing firms that knowingly deploy a common AI pricing engine — and who have either actual or constructive knowledge of that engine’s market-coordinating function — as participants in a concerted practice under the amended Section 3.
In the United States, the In re: RealPage, Inc. Rental Software Antitrust Litigation (No. II) represents the most consequential algorithmic collusion enforcement action to date. RealPage, a Texas-based property management software company, supplied algorithmic rent-setting software to competing landlords across major American cities. The software ingested non-public, competitively sensitive rental pricing data shared among landlords and generated pricing recommendations that had the systematic effect of elevating rents above competitive levels across multiple markets. The Department of Justice filed suit in a federal court in North Carolina, characterising the arrangement as a hub-and-spoke conspiracy in which RealPage (the hub) facilitated price coordination among competing landlords (the spokes) through the medium of its shared algorithmic platform. In November 2025, the DOJ and RealPage reached a settlement under which RealPage agreed to cease using real-time, non-public, competitively sensitive data for pricing recommendations a significant enforcement outcome that fell short of the structural remedies sought by private plaintiffs in the parallel multidistrict litigation, which by October 2025 had yielded preliminary class action settlements totalling $141.8 million with twenty-six defendant landlords. The RealPage litigation is significant for India’s purposes because it validates the hub-and-spoke framework as an enforcement mechanism against shared AI pricing systems, while simultaneously illustrating the evidentiary difficulty of proving the requisite agreement or concerted practice — a difficulty that resulted in a settlement without any admission of liability on RealPage’s part.
IV. The Evidentiary Void: Detection, Proof, and the Opacity of Algorithmic Coordination
The most practically formidable challenge in enforcing competition law against algorithmic collusion is not doctrinal but evidentiary. Traditional cartel investigations rely on direct evidence communications, emails, meeting minutes, leniency applications and circumstantial evidence parallel conduct combined with “plus factors” such as motive, opportunity, and market structure consistent with coordination. Algorithmic collusion leaves no such paper trail. Where competing firms’ AI systems reach coordinated pricing outcomes through independent machine learning, there are, by definition, no communications to intercept, no meetings to investigate, and no leniency applicants to approach, because no human actor is aware that coordination is occurring at the algorithmic level.
The CCI’s 2025 Market Study on AI and Competition expressly acknowledges this enforcement challenge, noting that self-learning and signalling algorithms may produce coordinated pricing behaviour that makes “detection of market collusion tougher.” Signalling algorithms — AI systems that monitor competitors’ prices in real time and adjust prices rapidly in response, effectively communicating competitive intentions through price movements rather than words — are particularly difficult to distinguish from vigorous competitive behaviour, since both competitive and coordinated pricing in dynamic markets produce rapid price adjustment in response to market signals. The evidentiary burden imposed by Section 3(3) of the Competition Act, read in the context of Section 19(3)’s multi-factor AAEC analysis, requires the CCI to establish not merely that prices were coordinated but that the coordination arose from an agreement or concerted practice within the meaning of Section 2(b). Demonstrating this in the context of autonomous algorithmic behaviour requires the CCI to move beyond price-outcome analysis to algorithmic process analysis — to examine the architecture of the AI systems involved, the data inputs they processed, the feedback mechanisms through which they adapted, and the market-structural conditions that made coordinated outcomes possible and profitable.
This form of technical investigation is largely beyond the current forensic capacity of the CCI, which, unlike competition authorities such as the UK’s Competition and Markets Authority or the French Autorité de la concurrence, does not possess a specialist algorithmic investigation unit or the technical infrastructure to conduct deep algorithmic audits of commercially deployed AI systems. The CCI’s Market Study recommends the establishment of a self-audit framework for algorithmic risks and proportionate transparency requirements to reduce information asymmetry, but these are advocacy tools rather than enforcement mechanisms, and their efficacy in the absence of compulsory algorithmic disclosure obligations and independent audit powers is fundamentally limited. The leniency plus framework introduced by the 2023 Amendment — which enables a cartel participant who is also a leniency applicant to disclose the existence of a second cartel and receive reduced penalties for both — is theoretically available as a detection mechanism for algorithmic cartel cases, but its utility depends on the existence of a human actor with knowledge of the coordination, which in a fully autonomous algorithmic collusion scenario may not exist.
V. Towards a Functional Accountability Framework: Policy Prescriptions for Indian Competition Law
The analytical weight of the foregoing sections converges upon a single conclusion: India’s competition law framework, as amended in 2023, is a necessary but not sufficient response to the challenge of AI-driven price coordination. The hub-and-spoke provisions of the amended Section 3(3) extend the reach of cartel liability to AI platform providers in a manner consistent with international best practices, and the Eturas knowledge-based presumption offers a doctrinal template that Indian courts and the CCI should affirmatively adopt in their interpretation of the amended provision. However, the existential challenge of AI-driven collusion in India will not be resolved through doctrinal interpretation alone — it demands a structural policy response built on at least five foundational pillars.
First, India must enact a statutory presumption of algorithmic coordinated conduct, modelled on the Eturas framework but calibrated to the Indian context, which provides that where two or more competing enterprises knowingly deploy a shared or substantially similar AI pricing engine, and where the output of that engine produces sustained supra-competitive pricing across the relevant market, an irrebuttable presumption of concerted practice shall arise under Section 3(3), subject only to the enterprise demonstrating that it took active steps to override or publicly disavow the algorithmic coordination. This reversal of the evidentiary burden from the CCI to the enterprise is both normatively justified the enterprise is in a uniquely privileged position to demonstrate the architecture and intent of its own algorithmic systems and practically necessary, given the enforcement capacity constraints that make affirmative proof of algorithmic agreement effectively impossible for a regulatory authority without deep technical access.
Second, the Competition Act must be amended to introduce a distinct sub-category of prohibited agreement denominated “algorithmic coordination” under a new Section 3(3)(e), which would provide that any arrangement, whether express or emergent, by which competing enterprises use AI-mediated systems to achieve price coordination producing an appreciable adverse effect on competition shall be deemed a per se violation of Section 3, irrespective of the presence or absence of a human agreement. This amendment would resolve the doctrinal impasse created by the agreement requirement in Section 2(b) by treating the algorithmic output — rather than the human intent behind it as the locus of liability. Third, the CCI must be endowed with the technical capacity and statutory authority to conduct compulsory algorithmic audits of AI systems deployed by enterprises under investigation for Section 3 violations, including the power to require the production of model architecture documentation, training data inventories, pricing output histories, and algorithmic decision logs. Fourth, mandatory algorithmic transparency obligations must be imposed upon enterprises deploying AI pricing systems in high-risk sectors defined by market concentration, consumer vulnerability, and pricing sensitivity — requiring them to register their algorithmic pricing systems with the CCI, disclose the nature of their training data, and demonstrate through a conformity assessment that the system does not produce coordinated pricing outcomes. Fifth, the CCI must establish a dedicated AI and Digital Markets unit, staffed with economists, data scientists, and algorithmic forensic specialists, with the investigative mandate and technical infrastructure to detect, analyse, and prosecute AI-driven competitive harms. India’s AI market is projected to grow to nearly $32 billion by 2031, and the cost of regulatory inaction — measured in suppressed consumer welfare, foreclosed market entry, and entrenched platform dominance — will compound with every year in which the legal framework lags behind the technological reality.
ConclusionThe question posed in the title of this article — whether India’s antitrust framework is equipped to detect and penalise AI-driven price coordination — admits of an honest answer that is simultaneously reassuring and alarming. It is reassuring because the Competition (Amendment) Act, 2023 demonstrates legislative awareness of the hub-and-spoke architecture of algorithmic collusion and has taken a meaningful doctrinal step toward extending cartel liability to AI platform providers who facilitate competitive coordination. It is alarming because the most dangerous and prevalent form of algorithmic collusion — the autonomous, emergent, machine-level coordination that leaves no paper trail, requires no human agreement, and produces cartel-like outcomes without any identifiable human conspiracy — remains entirely beyond the practical reach of India’s current enforcement architecture. The Samir Agrawal decision has established a restrictive doctrinal baseline that the 2023 Amendment has softened but not dismantled. The Eturas and RealPage frameworks offer instructive international models that Indian competition jurisprudence has not yet absorbed. The CCI’s own market study has diagnosed the problem with commendable clarity but prescribed solutions — self-audits, advocacy, capacity-building — that fall considerably short of the structural regulatory intervention the problem demands. The time for India to move from regulatory diagnosis to statutory prescription is not a matter of administrative preference; it is a matter of constitutional obligation to the millions of consumers whose welfare is at stake in the digital markets that algorithmic systems are rapidly reshaping.