Rethinking Cartels in the Age of Algorithms

Abstract

The digital age has introduced new challenges to competition law with the rise of artificial intelligence and advanced algorithms. In India, the Competition Act struggles to tackle AI-driven collusion and algorithmic cartels. Traditional legal frameworks rely on clear human agreements, but AI can enable tacit collusion without direct communication. Self-learning algorithms that find collusive strategies independently make it hard to prove intent or assign liability. The Competition Commission of India also faces a technical expertise gap, hindering investigations. Distinguishing between fair competition and anti-competitive conduct in algorithmic markets adds further complexity. To address these issues, regulators should reinterpret laws, strengthen technical capacity, study market trends, adapt enforcement tools, and collaborate internationally. As AI advances, a proactive and informed regulatory approach is vital to ensure fair competition in the digital economy.

Introduction

The advent of artificial intelligence[1] and advanced algorithms[2] has transformed global economies, fostering unprecedented efficiencies and innovations. However, this technological leap also presents novel challenges to established legal frameworks, particularly in the realm of competition law. One of the most significant concerns is the potential for AI-facilitated collusion, often called algorithmic cartels.[3] As algorithms increasingly determine pricing, supply, and other strategic decisions, the traditional understanding of cartel behavior, which relies on explicit agreements between human actors, is rapidly eroding. This article delves into how AI redefines the landscape of anti-competitive agreements, examines the implications for Section 3 of the Competition Act, 2002[4] (hereinafter ‘the Act’), and proposes a forward-looking approach for Indian market regulators to effectively tackle the challenge of algorithmic collusion in the digital age.

Cartels under Indian Competition Law

Under Indian competition law, cartels are considered the most egregious form of anti-competitive conduct. Section 3 of the Act specifically prohibits any agreement among enterprises or persons that causes or is likely to cause an appreciable adverse effect on competition (AAEC) within India. This includes agreements to directly or indirectly determine purchase or sale prices, limit or control production or supply, allocate markets, or engage in bid rigging. For the Competition Commission of India (CCI) to establish a cartel, the primary requirement is the existence of an “agreement.”[5] This AI-facilitated anti-competitive practice poses a severe challenge concerning legality, detectability, and liability.[6]

The CCI has, over the years, adopted a pragmatic approach, recognizing that direct evidence of an agreement is rare. It often relies on circumstantial evidence, such as parallel conduct combined with Plus Factors (e.g., structural links), to infer the existence of an agreement. However, the foundational premise remains that human actors arrange or participate in a coordinated scheme. This traditional framework faces significant strain when confronted with the autonomous and seemingly ‘agreement-less’ nature of algorithmic coordination.

How AI Alters Cartel Behaviour

The integration of AI into business operations has introduced new paradigms of anti-competitive agreements, fundamentally altering the dynamics of cartel behavior. These range from algorithms that merely assist traditional cartels to sophisticated self-learning algorithms that independently achieve supra-competitive outcomes.

Tacit Collusion through Algorithms: One primary concern is tacit collusion through algorithms. Unlike explicit human agreements, algorithms can learn from market data, including competitors’ pricing strategies, and adapt their pricing to converge on a collusive outcome. For instance, if multiple firms use similar dynamic pricing algorithms, these algorithms might independently raise prices in response to competitors’ price hikes, leading to an equilibrium that mimics a cartel, without any direct communication or explicit agreement between human operators. The economic literature suggests that machine-learning pricing algorithms can maximize profits by cooperating with other algorithms, though competition authorities have yet to identify any instances of this.[7]

Hub-and-Spoke Cartels: Another manifestation is the hub-and-spoke cartel, where a central entity or algorithm facilitates coordination. This can occur when a dominant online platform provides a shared pricing algorithm to various sellers or when a common third-party software algorithm sets prices for multiple competing firms. Here, the ‘spokes’ (firms) might not directly interact with each other, but the ‘hub’ (algorithm, platform, or common entity) acts as a conduit for coordination, potentially leading to collusive outcomes. The challenge lies in determining whether the hub actively facilitates collusion or whether parallel conduct is merely a byproduct of the algorithmic design. In Samir Agrawal v. ANI Technologies Pvt. Ltd.,[8] a case involving Ola and Uber, the informant claimed that the platforms’ pricing algorithms facilitated collusion among drivers. The CCI rejected the claim, stating that drivers’ independent action, even when subject to common pricing rules, did not constitute concerted practice.

Self-Learning Algorithms: Furthermore, the emergence of self-learning algorithms complicates matters. These advanced AI systems can dynamically adjust their strategies based on real-time market conditions and competitor actions through machine learning. They can independently discover optimal collusive strategies, making it difficult to pinpoint human intent or even awareness of the collusive outcome. The absence of mens rea[9] or deliberate human arrangement makes proving an “agreement” under the existing legal frameworks exceedingly difficult.

International Jurisprudence and Policy Trends

Jurisdictions worldwide are grappling with the complexities of algorithmic collusion and adapting their competition law frameworks. The European Union and the United States have actively explored this area through extensive reports, workshops, and policy papers.

The European Commission, for instance, has acknowledged the hub-and-spoke cartel. Guidelines often emphasize that the transmission of information through a common vertical player can lead to concerted practice among horizontal competitors, but only if all parties are aware of the intended coordination.[10]

In the United States, the Department of Justice and Federal Trade Commission have signaled their intent to scrutinize algorithmic pricing practices. They recognize the difficulty in proving intent but suggest that firms could be held liable if they knowingly use algorithms to coordinate prices.[11]

The OECD has published several detailed reports outlining various types of algorithmic collusion and proposed policy recommendations, including adapting legal definitions and strengthening investigative tools.[12]

These international experiences highlight a common thread: adapting existing laws through interpretation, focusing on the responsibility of firms for the competitive outcomes of their chosen algorithms, and recognizing the need for specialized technical expertise within enforcement agencies.

Enforcement Challenges in India

The challenges for the CCI in tackling algorithmic collusion are multifaceted, stemming from both legal and technical limitations.

Proving an “Agreement”: First, the core hurdle lies in proving an “agreement” under Section 3 of the Act. In cases of tacit collusion through algorithms or truly autonomous self-learning algorithms, there may be no direct communication or explicit understanding between firms. The ‘meeting of minds’ concept, which is central to traditional cartel enforcement, becomes elusive. How does one infer an agreement when it is the algorithms, not directly the humans, that converge on a collusive outcome? This absence of mens rea or conscious parallel behavior makes it extremely difficult for the CCI to establish a contravention.

Technical Expertise Gaps: Second, the CCI faces significant technical expertise gaps. Investigating algorithmic cartels requires a deep understanding of data science, machine learning models, and complex software. Identifying whether parallel pricing is a natural market outcome or a result of algorithmic coordination requires sophisticated analytical tools and a workforce proficient in digital forensics and algorithmic analysis. The sheer volume and complexity of data generated by algorithmic pricing systems pose considerable challenges in terms of data access, analysis, and interpretation, often requiring access to proprietary algorithms and vast datasets.

Attribution of Liability: Third, identifying the responsible party is complex. Is it the firm that deploys the algorithm, the algorithm designer, or the software vendor? Should the liability extend to the algorithm itself if it autonomously initiates collusive behavior? Indian law, like many others, traditionally focuses on human entities and their intentions. Navigating these attribution challenges requires nuanced legal and economic understanding.

Distinguishing Legitimate Competition: Finally, distinguishing between legitimate competitive dynamic pricing and collusive algorithmic behavior is a formidable task. In competitive markets, firms often react quickly to competitors’ price changes, which can lead to rapid price adjustments and parallel behavior. Algorithms only amplify speed and scale. The CCI must develop sophisticated methodologies to differentiate between pro-competitive responses and anti-competitive coordination facilitated by algorithms.

The Way Forward for Indian Regulators

Addressing the challenge of algorithmic collusion requires a multi-pronged proactive approach from the CCI and Indian lawmakers.

1. Legal Reinterpretation and Guidance: First, a reinterpretation or potential amendment of Section 3 of the Act may be necessary. While a full legislative overhaul might be premature, the CCI could issue detailed guidance notes or advisories clarifying its stance on algorithmic pricing and emphasizing that firms remain accountable for the competitive outcomes of the algorithms they deploy. This could include a broader interpretation of “agreement” to encompass scenarios of consciously parallel algorithmic behavior.

2. Capacity Building: Second, significant capacity building within the CCI is paramount. This includes recruiting and training economists, data scientists, and forensic experts who are proficient in AI and machine learning. Developing in-house capabilities to analyze proprietary algorithms, interpret complex datasets, and understand algorithmic design choices will be critical for effective enforcement. Collaboration with academic institutions and technology companies could also bridge this expertise gap.

3. Proactive Market Studies: Third, the CCI should proactively undertake market studies in digitally intensive sectors. This will help identify markets susceptible to algorithmic collusion, understand prevailing algorithmic practices, and assess potential risks before overt anti-competitive effects occur. Such studies can inform policy interventions and help develop nuanced understandings of digital markets.

4. Adapting Enforcement Tools: Fourth, adapting enforcement tools, such as leniency programs, is crucial. Firms may be incentivized to report if their algorithms inadvertently lead to collusion, providing valuable insights into how these systems operate. However, this raises questions about who receives leniency (the firm, AI developer, or employee responsible for deployment).

5. Judicial Consistency: Fifth, the appellate bodies must adopt a more consistent approach to evaluating indirect collusion. Decisions such as Samir Agrawal reflect a conservative stance on what constitutes a concerted practice. While a cautious approach is understandable, a rigid insistence on direct evidence may defeat the purpose of competition law in technology-driven markets. Courts should acknowledge the changing nature of market interaction and permit the use of circumstantial and structural evidence to support enforcement actions.[13]

6. International Cooperation: Finally, fostering international cooperation with leading competition authorities will enable the CCI to learn from their experiences, share best practices, and potentially collaborate on cross-border investigations involving complex algorithmic cartels. This global exchange of knowledge is vital, given the borderless nature of digital markets.

Conclusion

The digital age, driven by artificial intelligence and complex algorithms, presents serious challenges to traditional competition enforcement under Section 3 of the Act. One of the key difficulties lies in establishing the existence of an agreement or intent in cases involving algorithmic collusion. Self-learning algorithms, tacit coordination without direct communication, and hub-and-spoke cartel arrangements make it harder to detect and prove anti-competitive behavior using conventional tools. In response, the CCI must adopt a more technologically informed, proactive, and flexible approach. These steps are essential to ensure that the benefits of innovation are realized while upholding fair competition in India’s rapidly transforming digital economy.

References

[1] “Artificial Intelligence,” HarperCollins Publishers Ltd, https://www.collinsdictionary.com/dictionary/english/artificial-intelligence (accessed July 11, 2024).
[2] “Algorithm,” HarperCollins Publishers Ltd, https://www.collinsdictionary.com/dictionary/english/algorithm (accessed July 11, 2024).
[3] “Better Finance,” BETTER FINANCE, https://betterfinance.eu/article/algorithm-cartels/ (accessed July 11, 2024).
[4] The Competition Act, 2002, § 3.
[5] The Competition Act, 2002, § 2(b).
[6] “Abuse of Dominance in Digital Markets,” Organisation for Economic Co-Operation and Development (OECD), 2021, https://doi.org/10.1787/4c36b455-en (accessed July 10, 2024).
[7] Emilio Calvano et al., “Artificial Intelligence, Algorithmic Pricing, and Collusion,” 110 American Economic Review 3267 (2020).
[8] Samir Agrawal v. ANI Technologies Pvt. Ltd., Case No. 37 of 2018, Competition Commission of India, https://www.cci.gov.in/antitrust/orders/details/228/0 (accessed July 10, 2024).
[9] “Mens Rea,” LexisNexis, https://www.lexisnexis.co.uk/legal/glossary/mens-rea (accessed July 10, 2024).
[10] Ijllr Journal, “Cartels in The Algorithmic Age: India’s Legal Framework For Hub-And-Spoke Collusion,” IJLLR Journal, July 2, 2025, https://www.ijllr.com/post/cartels-in-the-algorithmic-age-india-s-legal-framework-for-hub-and-spoke-collusion (accessed July 10, 2024).
[11] “FTC, DOJ, and International Enforcers Issue Joint Statement on AI Competition Issues,” Federal Trade Commission, July 22, 2024, https://www.ftc.gov/news-events/news/press-releases/2024/07/ftc-doj-international-enforcers-issue-joint-statement-ai-competition-issues (accessed July 11, 2024).
[12] OECD, “Algorithms and Collusion: Competition Policy in the Digital Age,” 2017, http://www.oecd.org/competition/algorithms-collusion-competition-policy-in-the-digital-age.htm.
[13] Ijllr Journal, “Cartels in The Algorithmic Age: India’s Legal Framework For Hub-And-Spoke Collusion,” IJLLR Journal, July 2, 2025, https://www.ijllr.com/post/cartels-in-the-algorithmic-age-india-s-legal-framework-for-hub-and-spoke-collusion (accessed July 10, 2024).