Competing Algorithms And Collusive Outcomes - Rethinking Antitrust Liability In India
Author: LawSphereX
Category: General Law
Published:
Updated:
COMPETING ALGORITHMS AND COLLUSIVE
OUTCOMES: RETHINKING ANTITRUST LIABILITY IN INDIA Divyansh Sonwani, B.A. LL.B. (Hons.), Law College Dehradun, Uttaranchal University Dehradun, Uttarakhand, India
ABSTRACT
The emergence of AI-driven pricing algorithms has fundamentally reshaped competitive markets, creating efficiencies while simultaneously raising concerns of algorithmic collusion. Unlike traditional cartels, algorithmic coordination may occur without explicit communication, enabling firms to reach supracompetitive outcomes through automated and self-learning systems. This article examines whether India’s competition law framework, particularly Section 3 of the Competition Act, 2002, is equipped to address such technologically mediated conduct. By analysing Indian jurisprudence and drawing comparative insights from the European Union, the United Kingdom, and the United States, the article highlights the doctrinal limitations of an enforcement regime rooted in concepts of agreement and human intent.
It argues for a more adaptive approach that emphasizes foreseeability, algorithmic design, and corporate oversight in assessing antitrust liability. The article concludes that modernizing competition law to address algorithm-enabled coordination is essential to preserving market integrity while fostering technological innovation.
Introduction
The growing deployment of AI-driven pricing tools has quietly but decisively altered how markets function. Prices can now be set or nudged instantaneously, recalibrated in response to vast streams of data that no human decision-maker could realistically process in real time. This technological shift undoubtedly promises efficiencies. Yet it also opens the door to a more elusive antitrust problem. What has come to be described as algorithmic collusion arises when independent pricing algorithms, acting without any express coordination, drift toward supracompetitive prices that closely resemble cartel outcomes. The mechanics are subtle but potent. Pricing algorithms that continuously track rivals’ prices and respond by matching increases or punishing undercutting can, over time, stabilise market prices at artificially high levels.
Deviation becomes irrational, not because firms have agreed not to compete, but because the algorithmic environment itself makes defection costly and fleeting. As European competition authorities have noted, such systems increase market transparency, detect deviations almost instantly, and render punishment mechanisms more effective turning algorithms into facilitators rather than mere observers of collusion. Indian regulators have not been blind to this risk. The Competition Commission of India has cautioned that self-learning pricing systems may generate cartel-like outcomes “without communication,” unsettling the traditional foundations of antitrust analysis. The difficulty, of course, lies in doctrine.
Section 3 of the Competition Act is premised on concepts such as “agreement,” “action in concert,” and concerted practice ideas historically anchored in some form of human meeting of minds. Algorithmic coordination strains these categories, making proof of collusion both conceptually and evidentially complex. Against this backdrop, the article examines whether India’s competition law is equipped to confront these challenges. It revisits judicial reasoning in cases such as Samir Agrawal v. ANI Technologies, reflects on cartel jurisprudence in sectors like aviation, and places India’s approach alongside emerging policy responses in the EU, the UK, and the United States.
Ultimately, it argues for a recalibration of liability one that looks beyond express agreement and instead foregrounds foreseeability, algorithmic design choices, and the degree of corporate control exercised over pricing systems. Without clearer guidance and sharper investigative tools, algorithmic coordination may continue to operate in the uncomfortable space between innovation and illegality.
Indian Competition Law: Agreements and “Concerted” Practices The Competition Act, 2002 sets its face firmly against coordination that distorts markets. Section 3(1) prohibits agreements that cause, or are likely to cause, an appreciable adverse effect on competition, while Section 3(3) goes further, singling out cartels and other horizontal restraints price-fixing, output limitation, market allocation as especially suspect. What is striking, and often underappreciated, is how broadly the Act frames the very idea of an “agreement. ” It is not confined to contracts or explicit bargains. The statute tells us, quite deliberately, that an agreement includes any arrangement, understanding, or action in concert. That breadth matters. It means that coordination need not be formal, written, or even openly articulated to attract scrutiny.
1 Tacit alignment, if it reflects a shared commercial design rather than mere coincidence, can fall within the Act’s reach. Once cartel conduct is established, the law goes further still: such activity is presumed to have an appreciable adverse effect on competition, effectively shifting the burden onto the parties to demonstrate redeeming, procompetitive justifications. Section 3(4) extends the logic to certain vertical restraints as well, such as resale price maintenance, reinforcing the Act’s scepticism toward practices that quietly dampen rivalry. Although Indian competition law does not explicitly deploy the phrase “concerted practice,” the concept is functionally present through the idea of “action in concert. ” In principle, this language is capacious enough to accommodate new forms of coordination, including those mediated by technology.
If rival firms knowingly deploy compatible pricing algorithms, or adopt reactive AI systems in parallel with a shared understanding of how those systems will behave, their conduct could plausibly be characterised as acting in concert even in the absence of emails, meetings, or signed agreements. And yet, there is an unease here. For all its textual elasticity, the Act’s conceptual foundations remain rooted in assumptions of human intent and mutual awareness. Algorithms complicate that picture. 2 They act, learn, and respond at speeds and in ways that strain traditional notions of agreement. The law, on paper, may be flexible enough to stretch around algorithmic coordination. Whether its underlying logic can do so 1 Aman Avinav, Shivani Sinha & Rahul Deodhar, Safeguarding Sustainable Investments: Green Audits as the New Baseline for Due Diligence, Lexology (Nov. 4, 2025), https://www.
lexology. com/library/detail. aspx? g=ac1ed1cf-5ef5-4a24-9a5c-a6f49686ec74.
2 Addressing Algorithmic Collusion in Indian Competition Law, Centre for Business & Commercial Law,
National Law Institute University (CBCL NLIU), available at https://cbcl. nliu. ac. in/competition-law/addressingalgorithmic-collusion-in-indian-competition-law/ without distortion is a question that remains unresolved. Indian Precedents: Algorithms, Aggregators, and Aviation Cartels Indian competition jurisprudence has, so far, had limited occasion to grapple directly with algorithmic collusion as a distinct phenomenon. That said, a handful of decisions offer useful clues as to how Section 3 has been approached in contexts involving pricing software and parallel conduct in concentrated markets. The most frequently cited and still the most instructive is Samir Agrawal v. ANI Technologies Private Limited & Ors. , decided in November 2018.
3 The dispute arose from a complaint by a taxi driver against app-based cab aggregators, notably Ola and Uber, who were accused of using algorithmic pricing to fix fares across their platforms. According to the complainant, the fare-setting algorithm left driverpartners with no real autonomy on price, effectively eliminating competition among them. This, he argued, amounted to a hub-and-spoke cartel: the platform functioned as the coordinating hub, while the drivers compelled to accept algorithmically determined prices formed the spokes. The legal framing was telling. Rather than alleging a classic horizontal cartel, the complainant characterised the arrangement as resale price maintenance under Section 3(4)(e) of the Competition Act.
Because drivers could not undercut or deviate from the price generated by the app’s algorithm, the system, it was argued, imposed a vertical restraint that mirrored cartel-like outcomes. While the case did not ultimately turn on algorithmic collusion as such, it remains one of the earliest Indian attempts to translate concerns about automated pricing into the familiar grammar of competition law. The Commission was ultimately unpersuaded. It rejected the allegation that the drivers were party to any collusive arrangement, stressing that mere acquiescence to an algorithmically generated price does not, by itself, amount to agreement. For a hub-and-spoke cartel to exist, the spokes must do more than passively comply; they must either agree among themselves to fix prices or, at the very least, consciously decide to delegate pricing authority to the hub.
On the facts, there was no evidence of such a shared understanding. The drivers had not coordinated with one another, nor had they jointly resolved to be bound by the platform’s pricing logic. The resale price maintenance argument fared no better. The Commission noted that cab aggregators do not purchase transport services from drivers for onward resale. Instead, 3 Samir Agrawal v. ANI Technologies Pvt. Ltd. & Ors. , Case No. 37 of 2018, Competition Comm'n of India (Nov. 6, 2018), https://indiankanoon. org/doc/84896048/. the platform offers a composite service directly to passengers combining matching, routing, and payment facilitation. In the absence of a conventional resale chain, the doctrinal foundation for resale price maintenance simply did not exist.
The algorithm, the Commission observed, operated on the basis of inputs such as traffic conditions, location, and time parameters supplied and constrained by human design rather than as an autonomous price-fixing device. With no evidence of a human-driven arrangement or mutual understanding, the Commission concluded that Section 3 had not been contravened. The broader takeaway from Samir Agrawal is therefore a narrow one but an important one: algorithmic pricing, without proof of human coordination, does not offend Indian competition law as it presently stands. A similar pattern emerges from the Commission’s later inquiry into alleged cartelisation in the airline industry. 4 There, domestic airlines were accused of coordinating air cargo fuel surcharges through a shared third-party pricing tool.
Although information was pooled through a common system, the Commission found that final pricing decisions were still taken by airline personnel, informed by cost structures and commercial judgment. The software, in its view, facilitated price discovery; it did not independently determine or elevate prices. In the absence of evidence that the algorithm itself imposed supracompetitive fares, the case was closed. Read together, these decisions reveal a consistent judicial instinct. Indian competition law has, thus far, been willing to intervene where technology is used to implement or reinforce human collusion. What it has not yet confronted and perhaps is not doctrinally equipped to confront is coordination that emerges from self-executing algorithms without a provable meeting of human minds. The contrast with traditional cartel enforcement is instructive.
In more conventional cases, parallel pricing has often been enough to trigger an inference of collusion, especially when it unfolds with suspicious precision. A frequently cited example is the 2015 decision in which the Competition Commission penalised Jet Airways, IndiGo, and SpiceJet for simultaneously imposing identical cargo fuel surcharges. Faced with synchronised price hikes and the absence of any credible, cost-based explanation, the Commission concluded that “the only plausible reason” for the increase was collusion among the airlines. Parallel conduct, in that setting, was not treated as coincidence but as a tell-tale symptom of a shared commercial agenda. What made the difference, however, was context.
The price alignment did not exist in isolation; it was reinforced by evidence of coordination meetings, exchanges, and a pattern of conduct that 4 Rishabh Yadav & Shreyans Jain, Algorithmic Collusion: The CCI's Recognition and Regulatory Challenges in India, Lexology (Apr. 25, 2025), https://www. lexology. com/library/detail. aspx? g=69e69cbd-28d9-4cee-9228- 8154ae13d3a5. pointed toward an express or at least tacit understanding. 5 Human agency was visible, traceable, and inferable. That evidentiary scaffolding allowed the Commission to bridge the gap between parallel behaviour and unlawful agreement. Algorithmic collusion, by comparison, operates in a far murkier space. Prices may move in lockstep, sometimes with uncanny speed and consistency, yet without meetings, phone calls, or any recognisable “smoking gun.
” The coordination is embedded in code, learning models, and feedback loops rather than in human interaction. As a result, while the economic harm may look familiar, the legal proof does not. Under Section 3’s existing framework so deeply anchored in notions of human intent and mutual awareness this absence of overt coordination makes algorithmic collusion significantly harder to capture, even when its market effects resemble those of a classic cartel. Comparative Insights: EU, UK, and US Authorities Competition authorities across jurisdictions are increasingly alive to the complications introduced by algorithm-driven pricing, and their message has been fairly consistent: technology does not confer immunity.
The European Union, the United Kingdom, and the United States have all signalled sometimes explicitly that delegating pricing decisions to machines does not place firms beyond the reach of competition law. Nowhere is this clearer than in the European Union’s evolving position. As early as 2017, Margrethe Vestager cautioned that companies would not be permitted to “hide behind a computer program” if their automated systems were capable of collusion. 6 The warning was less rhetorical than it appeared. It reflected a growing regulatory discomfort with the idea that responsibility could dissolve simply because coordination was mediated by code rather than conversation. That instinct has since been formalised. The draft EU Horizontal Guidelines released in July 2023 squarely confront the role of pricing algorithms.
They make the position unmistakably clear: conduct that would be unlawful if carried out offline does not become lawful merely because it is executed by an algorithm. Firms cannot escape liability on the ground that prices were set by automated systems. An algorithm, in this framing, is treated as an agent operating under the firm’s control no different in principle from an employee, consultant, or outsourced decisionmaker. The Guidelines go further still. They recognise that coordination need not take the form
5 Xpress Industry Council of India v. Jet Airways, CUTS Int'l (2018).
6 EU Commissioner Warns Companies of Potentially Unlawful Use of Pricing Algorithms, K&L Gates (Mar.
30, 2017), https://www. klgates. com/EU-Commissioner-Warns-Companies-of-Potentially-Unlawful-Use-of- Pricing-Algorithms-03-30-2017. of explicit agreement. Where firms embed shared pricing rules into common algorithmic tools rules that, for instance, automatically match a rival’s lowest price and add a fixed margin such indirect coordination may still fall within the prohibition of Article 101(1). The absence of overt communication does not sanitise the outcome. What matters is the foreseeable effect: stabilised prices and softened competition. Taken together, the EU’s approach reflects a deliberate shift away from formalism. Algorithmic coordination, even when silent and decentralised, is treated as a genuine antitrust concern. The law looks past the machinery and fixes its gaze on control, design choices, and economic effects.
In that sense, the message is blunt but pragmatic: collusion does not cease to be collusion simply because it is written in code rather than spoken aloud. 7 The United Kingdom has moved in much the same direction, with the Competition and Markets Authority adopting a noticeably alert and at times pointed stance on algorithmic pricing. The CMA’s instincts were on display as early as 2016, when it fined two online sellers of posters and frames for using repricing software to give effect to an existing price-fixing arrangement. The software did not soften the infringement; it merely automated it. More recently, the Competition and Markets Authority has sharpened its warning.
In a blog post published in November 2024, the authority cautioned that the use of common or shared pricing algorithms can itself become a conduit for exchanging competitively sensitive information, allowing firms to drift sometimes quite comfortably into tacit price alignment. 8 The concern is not limited to explicit coordination. Even silent dependence on the same algorithmic logic can narrow competitive margins in ways the law is unwilling to ignore. What is particularly striking is the CMA’s emphasis on foreseeability. Firms are expected to think ahead. If it is reasonably foreseeable that a pricing tool might be drawing on confidential competitor data, or embedding assumptions that dampen rivalry, the firm may still be on the wrong side of the law. The message is unmistakable: responsibility does not evaporate simply because decisions are delegated to software.
This approach resembles a nuanced rule-of-reason analysis, one that looks closely at algorithmic design, incentives, and likely market effects. Companies are expected to police their own tools, to understand how they work, and to ensure that efficiency gains do not come bundled with forbidden coordination. The CMA’s conclusion is clear enough, even if its tone remains measured. Shared AI pricing tools are legally risky terrain, 7 Eur. Comm'n, Guidelines on the Applicability of Article 101 of the Treaty on the Functioning of the European Union to Horizontal Co-operation Agreements, C(2023) 3445 final (2023).
8 Competition & Markets Auth., Pricing Algorithms and Competition Law (Nov. 8, 2024).
and UK competition law, in the authority’s view, is fully capable of reaching tacit algorithmic collusion where the circumstances warrant it. In the United States, the conversation around algorithmic collusion has been neither abstract nor hypothetical. Enforcement agencies and courts have been forced, sometimes uncomfortably, to confront what happens when price coordination migrates from smoke-filled rooms to server rooms. The Department of Justice and the Federal Trade Commission have been unusually clear on first principles: Section 1 of the Sherman Act does not lose its force merely because coordination is executed by software rather than spoken aloud. That position came into sharp focus in Cornish-Adebiyi v. Caesars, where, in March 2024, the Federal Trade Commission and the United States Department of Justice jointly filed a Statement of Interest. 9 Their message was blunt.
Hotels cannot lawfully use pricing algorithms to coordinate room rates, and plaintiffs need not plead direct communications to establish an agreement. If competitors act “in concert” by adopting shared algorithmic pricing recommendations particularly through a common service provider that may suffice. Retaining nominal discretion over final prices does not cleanse the conduct. In practical terms, U. S. enforcers are treating algorithm-assisted price-fixing as no different from any other per se unlawful cartel. This stance is not new. As far back as 2015, the DOJ prosecuted a cartel of online poster sellers on Amazon, noting almost with a hint of incredulity that the conspirators had simply written code instructing their pricing software to implement their agreement. The medium changed; the offence did not. Yet U. S. courts have also drawn careful boundaries. In Gibson v.
Cendyn Group (Ninth Circuit, 2023), plaintiffs challenged independent hotels for using the same revenue-management software. The claim failed. The court found no allegation of an agreement to fix prices and no suggestion that the software shared confidential data among rivals. Mere parallel conduct, even when mediated by identical algorithms, was not enough. At the same time, the court left the door conspicuously ajar: had the hotels agreed among themselves to follow the software’s recommendations, the arrangement would have resembled a classic hub-and-spoke conspiracy per se unlawful under Section 1. The distinction is subtle but crucial. Independent use of tools is permissible; agreement to collude through them is not. Beyond the courtroom, the regulatory 9 FTC & U. S. Dep't of Just. File Statement of Interest in Hotel Room Algorithmic Price-Fixing Case, Fed. Trade Comm'n (Mar.
28, 2024), https://www. ftc. gov/news-events/news/press-releases/2024/03/ftc-doj-file-statementinterest-hotel-room-algorithmic-price-fixing-case. mood has been shifting steadily toward intervention. 10 Legislative proposals, such as the Preventing Algorithmic Collusion Act, reflect growing unease with pricing technologies that blur the line between competition and coordination. International bodies echo this concern. A
2025 survey by the Organisation for Economic Co-operation and Development notes that G7
competition authorities are actively studying algorithmic pricing through enforcement actions and market inquiries. Across jurisdictions from Japan to Canada to the European Union regulators have observed the same pattern: algorithms increase market transparency, accelerate interaction, and can make collusion both easier to reach and harder to disrupt. The OECD’s conclusion is telling. Pricing algorithms that facilitate coordination should be treated like any other cartel mechanism; what needs updating is not the law’s moral compass but its evidentiary toolkit. That sentiment surfaced again in a joint July 2024 statement by EU, UK, DOJ, and FTC authorities on generative AI, which warned explicitly of risks ranging from price-fixing to collusion and discriminatory pricing. Taken together, the trajectory is clear.
Across jurisdictions, the regulatory consensus is hardening around a simple idea: algorithmic assistance does not neutralize antitrust liability. What it does is complicate proof. And that, increasingly, is a challenge regulators appear willing perhaps even eager to meet. 11 Doctrinal Gaps: “Agreement” and Agency in India These comparative developments throw into sharp relief the fault lines in Indian competition law. Section 3 of the Competition Act continues to revolve around a familiar axis: the existence of an “agreement,” an understanding, or some form of concerted practice. That framework works tolerably well for old-fashioned cartels hotel rooms booked for meetings, phone calls placed at odd hours, trade association gatherings where prices somehow “stabilise. ” It begins to creak, however, when pricing decisions are delegated to autonomous systems.
Indian case law has consistently insisted on at least tacit coordination. Samir Agrawal is emblematic: drivers who merely accepted fares generated by an algorithm were not treated as having agreed to fix prices. Acceptance, the Commission suggested, is not the same as assent. This doctrinal insistence has consequences. Section 3(3)’s presumption of illegality for horizontal restraints price-fixing, output limitation, market allocation only comes into play once an agreement is 10 Recent Developments in Algorithmic Pricing: U. S. Appeals Court Weighs In, Hogan Lovells (Apr. 3, 2024), https://www. hoganlovells. com/en/publications/recent-developments-in-algorithmic-pricing-us-appeals-courtweighs-in. 11 Global Regulators Issue Joint Statement on Competition in the Generative AI Sector, Morgan Lewis (July 2024), shown.
12 Without proof of intent or mutual understanding, the presumption never even gets off the ground. In effect, Indian law is well equipped to catch explicit or quid pro quo collusion, even when implemented through software. What it struggles with is a quieter phenomenon: coordination that emerges from algorithmic interaction rather than being consciously designed. At a deeper level, the difficulty is conceptual. Competition law has long treated collusion as an act of will a convergence of minds, however fleeting or tacit. Self-learning algorithms unsettle that assumption. These systems can “discover” supracompetitive pricing strategies on their own, without any human architect ever instructing them to do so. As one commentator has asked, pointedly, what remains of collusion once its human authors disappear?
The Act, after all, was written with human actors in mind, not autonomous agents that learn, adapt, and respond in ways even their designers may not fully anticipate. The Supreme Court’s insistence, in Bharti Airtel Ltd. v. Competition Commission of India, that parallel conduct alone is insufficient that there must be evidence of a meeting of minds only sharpens the problem. In an oligopolistic market, similar AI models may converge on high prices while firms remain scrupulously independent. Under the current Section 3 framework, this looks like lawful parallelism, not illegal collusion. 13 Questions of liability and intent further complicate matters. If supracompetitive outcomes arise purely from code, should firms be sanctioned for failing to anticipate them? The law assumes control. Algorithms, particularly complex or opaque ones, resist that assumption.
Indian law lacks an intermediate concept something akin to the EU’s focus on object-based restrictions that would allow liability to attach where anticompetitive effects were reasonably foreseeable, even if not expressly intended. Without such a bridge, enforcement risks oscillating between two unsatisfactory extremes: no liability at all if firms plead ignorance, or strict liability for outcomes they genuinely could not predict. Scholars have described this as a zone of regulatory paralysis, where neither intent-based nor purely effects-based analysis sits comfortably. There is also a more prosaic, but no less serious, evidentiary problem. Traditional cartels leave trails meeting records, coordinated announcements, suspicious correspondence. Algorithmic coordination may leave none of that. Prices change because servers update.
Decisions are embedded in code, logs, or third-party platforms. Investigators may need to interrogate pricing 12 Samir Agrawal v. ANI Technologies Pvt. Ltd. & Ors. , Case No. 37 of 2018, Competition Comm'n of India (Nov. 6, 2018), https://indiankanoon. org/doc/84896048/. 13 Addressing Algorithmic Collusion in Indian Competition Law, Centre for Bus. & Com. L. , Nat'l L. Inst. Univ. , https://cbcl. nliu. ac. in/competition-law/addressing-algorithmic-collusion-in-indian-competition-law/ (last visited June 3, 2026). data, examine source code, or obtain records from software vendors. Indian enforcement practice has yet to articulate what, exactly, would count as sufficient algorithmic evidence. Is reliance on a common pricing tool enough to infer concerted practice? Samir Agrawal suggests not, absent more.
Yet lessons from the EU and the UK indicate that shared IT systems, consultants, or intermediaries can themselves become vehicles for coordination where they transmit sensitive information. The gap, then, is not merely technological. It is doctrinal, conceptual, and evidentiary all at once. And until those layers are addressed together, algorithmic collusion will continue to sit uneasily at the edges of India’s competition law framework. 14 Towards Rethinking Liability: Foreseeability, Design, and Control India cannot afford a posture of regulatory patience as algorithmic pricing tools spread quietly but rapidly across markets. The law must learn to recognise a new species of coordination one that does not announce itself through meetings or messages, yet predictably steers prices away from competition.
At the same time, any reform must tread carefully, preserving the undeniable efficiencies that well-designed algorithms can bring. The challenge, then, is not to ban the tool, but to discipline its use. One starting point lies in how responsibility is conceptualised. Drawing inspiration from European practice, Indian competition law could begin to treat algorithms as agents of the firm extensions of managerial decision-making rather than neutral black boxes. 15 If a company designs, trains, or configures an algorithm to shadow competitors’ prices or suppress undercutting, that design choice itself may evidence an “understanding” or “action in concert. ” Where rival firms license the same pricing software and deploy it in identical ways, a tribunal could plausibly infer that they have, in effect, agreed to follow a shared pricing logic.
Section 3’s deliberately broad language is not ill-suited to this task; it merely needs to be read with technological realism rather than doctrinal nostalgia. 16 Closely related is the question of foreseeability. Regulators might develop a “should have known” standard, shifting the inquiry away from proof of intent and toward reasonable 14 Competition & Markets Auth. , Pricing Algorithms and Competition Law: What You Need to Know (Nov. 8, 2024), https://competitionandmarkets. blog. gov. uk/2024/11/08/pricing-algorithms-and-competition-law-whatyou-need-to-know/. 15 Addressing Algorithmic Collusion in Indian Competition Law, Centre for Bus. & Com. L. , Nat'l L. Inst. Univ. , https://cbcl. nliu. ac. in/competition-law/addressing-algorithmic-collusion-in-indian-competition-law/ (last visited June 3, 2026). 16 PCompetition & Markets Auth.
, Pricing Algorithms and Competition Law: What You Need to Know (Nov. 8, 2024), https://competitionandmarkets. blog. gov. uk/2024/11/08/pricing-algorithms-and-competition-law-whatyou-need-to-know/. anticipation of outcomes. If firms could reasonably foresee that a pricing algorithm would stabilise prices above competitive levels, accountability should follow. This echoes approaches elsewhere in law, where liability often turns on foreseeable harm rather than subjective motive. The United Kingdom’s Competition and Markets Authority has already gestured in this direction, warning firms that advantages drawn from shared or mutualised pricing data even without explicit exchanges may breach competition rules. Indian regulators could send a similar signal: reliance on common AI pricing infrastructure is not a neutral act and carries inherent competitive risk.
Clear guidance would also make a difference. The Competition Commission of India could issue policy notes or formal guidance clarifying that familiar cartel principles apply with equal force to algorithmic pricing. Whether prices are fixed by handshake or by machine, the prohibition is the same. Such guidance could identify practical red flags sudden parallel price increases among firms using similar software, opaque third-party intermediaries, or algorithms that respond aggressively to minor deviations. The Commission’s ongoing market study on AI and competition is a promising beginning; its conclusions should not remain academic, but feed into operational standards that firms and investigators alike can rely upon. None of this is workable without stronger investigative capacity. Algorithmic coordination leaves different traces.
Instead of emails and calendars, there are pricing logs, training datasets, and lines of code. Competition authorities will need technical expertise data scientists, forensic analysts, and the ability to interrogate digital systems. Powers to demand algorithmic documentation, decision logs, and data-sharing records will be essential. In appropriate cases, scrutiny may even need to extend to software vendors or cloud-service providers that sit quietly at the centre of pricing ecosystems. The objective is simple, if technically demanding: to see how prices are actually being generated, and whether coordination is being embedded in architecture rather than agreement. Preventive compliance has a role to play as well. Firms should be encouraged perhaps eventually required to audit their own AI pricing systems.
Documenting algorithmic design choices, data inputs, and pricing outcomes can help businesses identify unintended convergence early and demonstrate due diligence. If such internal reviews become standard practice, they may not only reduce risk but also serve as mitigating evidence where enforcement questions arise. Attention must also turn to platforms and intermediaries. Where a digital platform supplies pricing tools to multiple competitors, it may function intentionally or not as a hub of coordination. Ride-hailing apps, e-commerce seller tools, or industry-wide analytics platforms can synchronise behaviour simply by standardising logic. Indian policy discussions have already flagged the idea of heightened duties for such intermediaries.
Holding platforms accountable for facilitating alignment would recognise the structural power they wield, even when they do not set prices themselves. In the longer run, legislative clarification may be unavoidable. Parliament could amend the Competition Act to explicitly acknowledge algorithmic collusion, whether by expanding the definition of “agreement” or by inserting provisions that recognise software-mediated coordination. Care would be needed safe harbours for genuinely independent algorithms are tempting but controversial but even modest amendments could signal that code is not a refuge from competition law. Finally, none of this can happen in isolation. Algorithmic pricing tools are global, and enforcement will increasingly require cross-border cooperation.
Coordination with foreign competition authorities, shared studies, and joint investigations can help India avoid reinventing the wheel. If a pricing software vendor raises red flags abroad, Indian users of the same system deserve scrutiny at home. At bottom, the problem is not that algorithms exist, but that they allow coordination without conversation. Traditional notions of collusion so dependent on proof of human intent begin to wobble in that reality. If Indian competition law is to remain effective, it must focus less on whether firms shook hands and more on whether they designed systems whose predictable effect was to soften competition. Accountability, in this new landscape, lies as much in architecture as in agreement.
17 Conclusion The spread of autonomous pricing algorithms forces competition law to confront an uncomfortable truth: the old grammar of collusion is no longer quite adequate. India’s Competition Act is, on its face, admirably broad and purposive, but the way it is read and applied still assumes conscious human coordination a meeting of minds, however tacit. That assumption begins to wobble when pricing decisions are outsourced to systems that learn, adapt, and converge without direct instruction. Left unattended, rival algorithms may exploit these doctrinal seams, to the detriment of consumers and competitive markets alike. What is required is not a wholesale abandonment of existing principles, but a recalibration of emphasis.
A more workable approach would treat algorithms as instruments of the firms that deploy them and focus on foreseeable collusive outcomes rather than elusive proof of intent. That shift 17 Addressing Algorithmic Collusion in Indian Competition Law, Centre for Bus. & Com. L. , Nat'l L. Inst. Univ. , https://cbcl. nliu. ac. in/competition-law/addressing-algorithmic-collusion-in-indian-competition-law/ (last visited June 3, 2026) would justify closer scrutiny of shared AI pricing platforms, clearer guidance to firms on what constitutes lawful algorithmic design and deployment, and a serious investment in digital investigative capacity. In other words, the law must become fluent in the technologies it seeks to regulate. International experience already points the way.
European regulators have been explicit that firms cannot shield themselves behind software, while UK enforcers repeatedly stress that businesses are expected to anticipate competition risks embedded in algorithmic tools. U. S. agencies, for their part, have made it equally clear that algorithms do not confer immunity from antitrust norms. There is no reason India’s Competition Commission of India cannot send a similar signal: algorithm-enabled collusion will be treated as collusion, full stop. Striking the right balance between innovation and vigilance will require regulators to look past surface-level compliance and into the architecture of pricing systems themselves. Firms should understand that feeding rival price data into AI models, or relying on common pricing software, may in substance amount to an unlawful agreement even if no one ever picked up the phone.
By aligning enforcement with the lived realities of AI-driven markets, India can ensure that its competition regime remains resilient, credible, and effective, even when it is algorithms, rather than people, doing the coordinating.