Smart Pricing or Secret Cartel? How Competition Law Draws the Line on Algorithms

Every major airline, hotel chain, e-commerce platform and ride-hailing service now prices algorithmically. The technology is efficient, lawful and commercially rational. Yet the same algorithms that optimize margins can, in the wrong configuration, transmit Commercially Sensitive Information (CSI) between competitors, enforce common pricing rules, or produce visible price convergence that attracts regulatory scrutiny.

The legal question is deceptively simple: when does automated pricing become collusion? Drawing on recent enforcement actions from India, the European Union (EU) and the United States of America (US), this article identifies the evidentiary line separating lawful algorithmic responsiveness from unlawful coordination and explains why businesses in India face a distinctive procedural risk that makes proactive compliance essential.

The Cartel Taxonomy: Messenger, Hub-and-Spoke, and Predictable Agent

Competition law requires some form of agreement or concerted action — whether under Section 3 of the Competition Act 2002 (Act), Article 101 TFEU1, or Section 1 of the Sherman Act.2 The challenge is applying that requirement when pricing decisions are delegated to code. Ezrachi and Stucke’s taxonomy structures the analysis.3

In the Messenger Model, humans make the unlawful agreement, and the algorithm merely implements it. In Topkins (US) competitors agreed to fix prices on Amazon Marketplace and configured pricing software to give effect to that arrangement.4 The code was the delivery mechanism, not the source of the conspiracy.

The Hub-and-Spoke Model is harder because competitors may not communicate directly — a platform or software provider sits between them. In Samir Agarwal v. ANI Technologies (India), the Competition Commission of India (CCI) held that a hub-and-spoke cartel requires spokes to use the hub to exchange sensitive information and that there must be an agreement to fix prices.5 In Eturas (EU, Court of Justice of the European Union (CJEU)), awareness of a platform-wide pricing restriction combined with failure to distance oneself could support tacit assent, but liability could not rest on a system message alone.6 The modern version is RealPage (US), where the allegation was that software used non-public, property-specific rent and occupancy data to generate pricing recommendations for competitors.7 The case was later settled with the Department of Justice, with RealPage undertaking not to use competitors’ CSI in rental pricing recommendations and agreeing to redesign its systems.8

In the Predictable Agent scenario, a firm independently deploys a profit-maximizing algorithm that learns to follow the market. Existing law struggles here because conscious parallelism or tacit coordination, without an agreement or concerted practice, ordinarily falls outside agreement-based cartel rules — even if the outcome is higher prices.9

CSI: The Evidentiary Linchpin

CSI is the decisive evidentiary issue. The CCI’s guidance and the EU Horizontal Guidelines both treat prices, costs, margins, capacity, market shares, customer terms and strategy as commercially sensitive, while excluding public, historic, aggregated or non-company-specific information.10 

Recent US cases illustrate the line:

·       In Gibson v. Cendyn (US) and Cornish-Adebiyi (US), the courts held that mere use of common pricing software, without exchange of confidential competitor information, did not plausibly allege collusion. 11The dismissal in Gibson has since been upheld by the Ninth Circuit.12

·       DAI v. SAS Institute (US) adds a further refinement to this logic by holding that it is not enough to allege that confidential information was fed into an algorithm unless the plaintiff can show that the information was pooled out in a way that reduced competitors’ uncertainty or aligned pricing decisions.13

·       In contrast, in Duffy v. Yardi (US), the court did not dismiss the investigation because plaintiffs alleged that lessors supplied commercially sensitive data, knew competitors were doing the same, and traded that data for algorithmic recommendations designed to raise rents.14

Algorithmic pricing is not unlawful in itself under EU law. But where a shared platform or tool transmits CSI to competitors by design, the EU's concerted practice doctrine does not require proof of intent — awareness of the practice, combined with failure to publicly distance oneself, suffices to presume participation. The following cases illustrate this pattern:

·       In Eturas, a travel platform capped discounts via a message to all agents; the CJEU held that agents aware of the cap who kept using the platform without distancing themselves could be found to have tacitly coordinated.

·       In Proptech, a shared listing service and CRM systems made each agency’s commission visible to rivals, discouraging undercutting.15

·       In Ageras, a platform generated “estimated market prices” and warned partners who bid below them.16

The fulcrum in each case is whether CSI was present and transmitted, not whether anyone intended to collude.

Read together, the test cuts both ways. Where an algorithm relies only on public or independently obtained data, collusion will generally not be made out. Where it is trained on competitors' non-public data, the CSI issue is squarely engaged. Even then, the authority must connect the data flow to a concerted practice — CSI is necessary but not sufficient.

Operationally, this calls for a three-level enquiry:

·       At the input level: what data feeds the algorithm?

·       At the model level: whether competitor data is pooled into a common training set?

·       At the output level: whether the resulting recommendations reduce uncertainty between competitors?

These varying US and EU standards have a doctrinal explanation. In the US, price-fixing can carry criminal liability, which is why courts insist on clear proof of an agreement or intent. In the EU, competition law is enforced through civil penalties, and the evidentiary threshold is lower — no proof of intent is required once awareness of a CSI-transmitting practice and non-distancing are shown. In India, the CCI applies a dual framework: for clear-cut cases such as public procurement cartels, the fiction of rebuttable presumption shifts the burden to defendants; in less obvious cases, the CCI assesses whether CSI was actually exchanged at the threshold stage, without needing to prove intent – a question left to detailed investigation.

Therefore, the two pertinent questions are: (1) is the information CSI?; and (2) was there awareness of a practice transmitting CSI, combined with continued participation? Given how new this technology is, businesses should expect close regulatory scrutiny of both questions.

Mimicking is not collusion - but the enforcement risk is real

Mimicking a competitor’s price or other business-critical aspects, without an agreement or CSI exchange, is not collusion. The CCI’s airline cases confirm that price parallelism and use of public competitor prices do not establish liability under Section 3(3) of the Act unless attributable to information exchange or other collusive conduct.17 In the Tyre cartel case, the CCI held that “without additional evidence i.e. ‘plus factors’, proof of conscious parallel behavior is not enough to establish a violation.”18 Portillo v. CoStar makes the same point in the algorithmic context: benchmarking outputs based on aggregated historical hotel performance data did not amount to coordination because they did not involve current individual pricing, real-time prices or pricing recommendations.19 Plus factors that may transform mimicry into coordination include reciprocal data contributions, knowledge that rivals contribute comparable non-public data, high adherence to recommendations, or technical barriers to deviation.

The enforcement risk in India, however, is more immediate than doctrine suggests. The CCI’s evidentiary standard is “preponderance of probabilities”, and as recognized by the CCI and the Indian Appellate Courts20, cartel evidence is largely circumstantial given the cloak-and-dagger nature of such activities. Under Section 3(3) of the Act, once an “agreement” — which includes any concerted action — is established even circumstantially, the onus shifts to the defendants, and there is a presumption of appreciable adverse effect on competition.

For businesses, the pressing concern is not a final adverse finding but a direction under Section 26(1) of the Act ordering investigation.21 For ordering an investigation (i.e., prima facie order), the threshold is ordinarily lower. In a sector where algorithms produce visible price convergence and firms use common software vendors, a regulator may direct an investigation even where the parallelism is better explained by independent algorithmic responsiveness.

Once an order directing an investigation is passed, parties face years of document production, depositions and regulatory engagement. This burden falls equally on an understaffed CCI. The result is a lose-lose outcome that warrants proactive compliance well before the regulator comes calling.

The Human Layer: Reconstructing Intent

Artificial Intelligence (AI) does not erase human responsibility. Where managers instruct an algorithm to implement a price-fixing agreement, intent is direct. Where they use a shared data pool, intent may be inferred from the decision to supply CSI, knowledge of competitors’ participation, or acceptance of recommendations. Intent should be reconstructed from contemporaneous evidence: product-design documents, vendor contracts, audit logs and communications with competitors or the platform. This is consistent with VM Remonts (EU, CJEU), which held that liability for a third-party tool requires control, knowledge, intention or acceptance of risk — not mere outsourcing.22

A Six-Step Framework for the CCI

The CCI should incorporate CSI analysis expressly into algorithmic cases through a six-step framework:

·       Identify the alleged agreement type.

·       Classify the information under CCI’s CSI categories.

·       Map the data flow.

·       Test the human layer — i.e., knowledge, acceptance, distancing and override rights.

·       Examine plus factors rather than relying on price parallelism alone.

·       Tailor remedies without banning beneficial algorithmic pricing.

Remedies should be proportionate: data-siloing obligations, mandatory firewalls, algorithmic audit requirements, or contractual restrictions on recommendation adherence. The CCI’s existing powers under Section 27 of the Act provide a basis, though further guidance is needed.

Beyond Section 27 of the Act, a longer-term solution lies in how the CCI designates and monitors AI systems. The CCI’s Market Study on AI and Competition identifies algorithmic coordinated conduct, AI-facilitated collusion, dynamic pricing and reduced transparency as specific concerns.23 The CCI Chairperson has reinforced this trajectory through dedicated sessions on AI and competition.24 An ongoing market study assesses the efficacy of ex-ante regulations in India.25 If enacted, such regulations could impose data-sharing restrictions, transparency obligations and algorithmic audit requirements addressing CSI flows before they mature into cartel conduct. The regulatory direction is clear: the CCI intends to understand AI systems as market infrastructure capable of facilitating coordination. Compliance must extend to the design and governance of algorithmic systems.

Practical Takeaways for Business

Businesses may lawfully:

·       use algorithms to automate pricing;

·       scrape public prices;

·       benchmark against aggregated market data; and

·       deploy tools trained on their own data.

Businesses should not:

·       feed competitor-specific CSI into a shared tool;

·       accept recommendations generated from rival non-public data;

·       agree to pricing rules through platform messages; or

·       design systems that make deviation impractical.

When procuring third-party pricing software, firms should insist on data-segregation, audit rights and contractual assurances that the vendor does not pool proprietary data across competing users, and should be prepared to publicly distance themselves from, or report, any platform-communicated pricing constraint of which they become aware.

The Algorithm Is the Medium, Not the Problem

While regulators do not require a new theory of cartel law to address algorithmic pricing, they do require guidance on how to navigate the existing evidentiary and doctrinal tools. The key analytical questions are:

·       The evidentiary question: Has CSI been exchanged through the algorithm?

·       The jurisdictional question: In the US, has an agreement or intent to coordinate been shown, notwithstanding the algorithmic medium? In the EU and India, has the business been aware of a CSI-transmitting practice and failed to distance itself from it, given that intent is not typically required?

·       The substantive question: Has independent decision-making been replaced by practical cooperation — whether through an agreement, a concerted practice, or a shared tool that removes competitive uncertainty?

Therefore, the competition-law problem is not AI as such. It is the replacement of independent judgement with a coordinated process — the algorithm is simply the medium.

• 1Consolidated Version of the Treaty on the Functioning of the European Union [2016] OJ C202/47, art 101.
• 2Sherman Act 1890, 15 USC § 1.
• 3Ariel Ezrachi and Maurice E Stucke, Virtual Competition: The Promise and Perils of the Algorithm-Driven Economy (Harvard University Press 2016).
• 4United States v Topkins, No CR 15-00201-WHO (ND Cal, Information filed 6 April 2015); Plea Agreement (ND Cal, 30 April 2015) https://www.justice.gov/atr/case-document/file/513586/dl and https://www.justice.gov/atr/case-document/file/628891/dl.
• 5Samir Agrawal v ANI Technologies Pvt Ltd and others, Case No 37 of 2018 https://www.cci.gov.in/images/antitrustorder/en/3720181652328966.pdf
• 6Eturas UAB and Others v Lietuvos Respublikos konkurencijos taryba (Case C-74/14) ECLI:EU:C:2016:42 https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:62014CJ0074
• 7United States v RealPage Inc, No 1:24-cv-00710 (MDNC, complaint filed 23 August 2024) https://www.justice.gov/archives/opa/pr/justice-department-sues-realpage-algorithmic-pricing-scheme-harms-millions-american-renters
• 8United States v RealPage, Inc., No. 1:24-cv-00710-WO-JLW, Proposed Final Judgment, United States District Court for the Middle District of North Carolina, filed 24 November 2025, available at: https://www.justice.gov/atr/media/1419451/dl?inline.
• 9OECD, Algorithms and Collusion: Competition Policy in the Digital Age, DAF/COMP(2017)4, 16 May 2017, available at: https://one.oecd.org/document/DAF/COMP(2017)4/en/pdf.
• 10Competition Commission of India, Frequently Asked Questions: Providing Guidance Through Advocacy (Competition Commission of India, May 2025) https://cci.gov.in/images/whatsnew/en/faq-book-english-compressed1747724324.pdf  ; European Commission, ‘Guidelines on the applicability of Article 101 of the Treaty on the Functioning of the European Union to horizontal co-operation agreements’ [2023] OJ C259/1 https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:52023XC0721(01).
• 11Gibson v Cendyn Group LLC, No 2:23-cv-00140-MMD-DJA, 2024 WL 2060260 (D Nev, 8 May 2024) https://law.justia.com/cases/federal/district-courts/nevada/nvdce/2%3A2023cv00140/160470/183/; Cornish-Adebiyi v Caesars Entertainment Inc, No 1:23-CV-02536-KMW-EAP, 2024 WL 4356188 (D NJ, 30 September 2024) https://law.justia.com/cases/federal/district-courts/new-jersey/njdce/1%3A2023cv02536/512272/139/ .
• 12US Court of Appeals for the Ninth Circuit, Gibson v Cendyn Group LLC, No 24-3576, 2025 WL 2371948, 15 August 2025, available at https://cdn.ca9.uscourts.gov/datastore/opinions/2025/08/15/24-3576.pdf.
• 13.DAI v SAS Institute Inc, No. 4:24‑cv‑02537‑JSW, 2025 WL 2078835, United States District Court for the Northern District of California, Order dated 18 July 2025, docket available at: https://storage.courtlistener.com/recap/gov.uscourts.cand.428726/gov.uscourts.cand.428726.137.0.pdf.
• 14 Duffy v Yardi Systems Inc, No 2:23-cv-01391-RSL, 758 F Supp 3d 1283 (WD Wash, 4 December 2024) https://law.justia.com/cases/federal/district-courts/washington/wawdce/2:2023cv01391/326049/187/.
• 15S/0003/20 – Proptech, Comisión Nacional De Los Mercados Y La Competencia, Decision dated 25 November 2021, available at: https://www.cnmc.es/sites/default/files/3831141.pdf
• 16Danish Competition Council, Ageras, Decision of 30 June 2020, Case No. 18/19827, available at: https://kfst.dk/media/ws5nbdtx/20200630-ageras-final-a.pdf
• 17In Re: Alleged Cartelization in the Airlines Industry, Suo Motu Case No 03 of 2015 (Competition Commission of India, 22 February 2021) https://www.cci.gov.in/images/antitrustorder/en/0320151652249082.pdf ; Ms Shikha Roy v Jet Airways (India) Ltd and others, Case No 32 of 2016 (Competition Commission of India, 3 June 2021) https://www.cci.gov.in/images/antitrustorder/en/3220161652248293.pdf ; Competition Act 2002, s 3(3).
• 18Apollo Tyres Limited v. Competition Commission of India and Others, Reference Case No. 08 of 2013, Order dated 31 August 2018, available at: https://www.cci.gov.in/images/antitrustorder/en/0820131652434997.pdf
• 19Portillo v CoStar Group Inc, No. 2:24‑cv‑00229‑RSL, 2025 WL 2495053, United States District Court for the Western District of Washington, Order dated 29 August 2025, available at: https://cases.justia.com/federal/district-courts/washington/wawdce/2:2024cv00229/331643/117/0.pdf?ts=1756568987
• 20Cement Manufactures’ Association (CMA) v Competition Commission of India & Ors, TA (AT) (Compt.) Nos 13 and 27 of 2017, along with connected appeals, National Company Law Appellate Tribunal, New Delhi, judgment dated 25 July 2018, available at: Competition Commission of India, Government of India; https://nclat.nic.in/sites/default/files/migration/upload/9924885005c514c82465bf.pdf.
• 21Competition Act 2002, s 26(1).
• 22SIA VM Remonts, formerly SIA DIV un KO, and Anr. v Konkurences padome and Anr. v SIA ‘Pārtikas kompānija’ (Case C-542/14) ECLI:EU:C:2016:578.
• 23Competition Commission of India, Market Study on Artificial Intelligence and Competition (2025) https://www.cci.gov.in/images/marketstudie/en/market-study-on-artificial-intelligence-and-competition1759752172.pdf
• 24Competition Commission of India, ‘Smt Ravneet Kaur, Chairperson, CCI chaired the Plenary Session on “Understanding the AI Market and Competition Landscape”’ https://www.cci.gov.in/events/Advocacy/details/1199
• 25Ministry of Corporate Affairs, Government of India, Request for Proposal for Engagement of an Agency to Conduct Market Study on Digital Competition Landscape in India and Review of the Draft Digital Competition Bill, 2024, RFP No. RFP-20251104, 3 November 2025, copy available at: https://www.medianama.com/wp-content/uploads/2025/11/RFP-20251104.pdf