The law on competition was constructed against human wrongs in a market. The essence of cartels is one that assumes the existence of people, executives in hotel rooms, exchanging messages, creating a meeting of minds. Yet what happens when those who do not meet, do not communicate and do not even have the intention of colluding yet the market prices go up, even, and markets silently compete away? This is no more a hypothetical issue. Even without human coordination, global regulators publicly recognized that self-learning pricing algorithms cause structural antitrust risks in 2024. In its competition policy discourse of recent, the OECD cautioned that reinforcement-learning algorithms have the capacity of uncovering and perpetuating supra-competitive pricing results by themselves.[1] About the same period, the European Commission, in releasing the EU AI Act, identified algorithmic pricing as a systemic market risk- but did not go further and establish a cartel liability framework.[2]
India is not insulated. Dynamics in the competitive pricing tools that can be fast and automated to market response are already being witnessed by the Competition Commission of India, in its e-commerce and telecom market studies. In the meantime, the Digital Competition Bill 2024 is an indication of transitioning to ex-ante regulation of digital markets- but it does not say much about autonomous algorithmic collusion.[3]
This is important since we are at the beginning of Algorithmic Collusion 2.0: a world in which self-learning AI systems, which solely aim to maximise profit, are able to replicate the results of a cartel with no human intervention.[4] These systems do not concur or negotiate or get to know the law of collusion, but they are able to learn that joint price suppression is the most lucrative equilibrium. Whether humans colluded or not, either way, is no longer the question--but whether competition law will survive in a machine-learning market. This blog discusses this fault line, and why the enforcement of antitrust law needs to shift to a more outcome-friendly approach before the algorithms of antitrust cartels turn into a new reality.
JURISPRUDENTIAL GAPS AND EMERGING CHALLENGES
Algorithmic collusions are AI-driven algorithms or tools that are widely used by firms in a competitive market to align or predict prices either through intentional programming or by autonomous self-learning. These algorithms are generally trained or computed by firms.[5] Then, according to the fed data, they determine the price variable in the market. However, the complexity intensifies in the context of self-learning systems, particularly those based on reinforcement learning, where algorithms continuously adjust their strategies through trial, error, and reward-based optimization. Self-learning algorithms occur when AI pricing tools independently discover that coordinating prices, rather than competing, maximizes profits, creating "cartel-like" behavior without explicit human agreements.[6] This concern has been reinforced by the Competition Commission of India (CCI), which notably warned about the possible collusion in the market by AI pricing systems even without human intent. However, Regulatory concern is no longer theoretical. At the 10th National Conference on Economics of Competition Law (March 2025), CCI Chairperson Ravneet Kaur warned that AI may enable "cartels without human communication" and "price coordination without explicit agreements," while also facilitating discriminatory pricing under the guise of dynamic pricing.[7] In addition, such algorithms not only hurt the market but also erode the distinction between explicit and tacit collusion.
The Problem of Liability Attribution
When self-learning algorithms rely solely on publicly available market data, liability attribution becomes more challenging. The challenge lies in distinguishing autonomous algorithmic adaptation from de facto collusive outcomes, an issue that represents an emerging frontier in competition law.[8]
The Crisis of Intent in Antitrust Doctrine
It brings out a more profound jurisprudential conflict with the systems of the current competition laws. The classic antitrust doctrine is founded on the fact that collusion needs to have some sort of agreement, communication or meeting of minds among the competing companies.[9] Nonetheless, as per algorithmic collusion 2.0, coordination can be achieved not by a human process but rather by machine-based optimization processes that come to learn that stable price coordination is more profitable.[10] Consequently, the legal instruments used to determine cartelization are becoming increasingly ineffective, as the element of willfulness, which is the core of enforcing cartels, may be missing or extremely difficult to prove. In this case, the companies can claim that the algorithm was simply acting rationally in response to market indicators, rather than engaging in any organized activity.
Evidentiary Burden and Algorithmic Opacity
The other challenge is in the responsibility attribution. The problem with self-learning algorithms is that they are based on complex layers of inputs and training models and adaptive decision-making systems, and it is difficult to decide who is liable, is the firm that is deploying the algorithm, the developers that created the algorithm or the algorithmic process itself.[11] The competition authorities are, therefore, put under an evidentiary burden to prove that a firm knowingly enabled collusive outcomes.[12] Such ambiguity undermines the role of competition law in deterrence, as companies can use algorithmic obscurity to dissociate themselves from potentially anti-competitive outcomes. Moreover, the current competition law jurisprudence is mainly reactive in nature, meant to respond to collusion when the communication or coordination has been proven. Nonetheless, algorithmic markets have the potential to reach speeds and scale such that a collusive equilibrium could be reached and sustained before regulatory forces can intervene.[13] Algorithms might enable coordinated pricing by using predictive analytics, real-time competitor tracking, and automatic retaliation mechanisms, which do not rely on the traditional signals used by competition authorities to identify cartels.
Speed, Scale and the Limits of Reactive Enforcement
It is based on these reasons that the current law system seems to be poorly equipped to address algorithmic collusion in its sophisticated form. Human intention, traceable communication, and recognizable decision-making structures are all assumed in the law, but self-learning algorithms undermine each of them. With the ongoing growth of algorithmic decision-making in digital markets, competition regulators might have to reconsider evidentiary principles, allocation of liability and regulatory instruments to successfully address the dangers of autonomous pricing mechanisms.[14] Failure to do so will keep increasing the disparity between technological capacity and law enforcement, thus making it possible to have an algorithm-driven coordination that is at least systematically beyond the application of the classical antitrust dogma.
CARTEL REGULATION AND THE EXPANDING CONCEPT OF AGREEMENT
The formation of cartel in the markets with the intention to limit, control or attempt to control the production, distribution, sale or price of, or, trade in goods or provision of services as defined under section 2(c) of the competition act of India is not a new phenomenon, but has been part of doing business since the antiquity.[15] But the need to regulate cartels solidified in the late nineteenth century in the excess of trust formation led to the introduction of modern competition law with the passage of the Sherman Act of US. Section 1 of the act prohibit contracts and combinations, or conspiracies that are hampering trade with the emphasis put on the need to have a meeting of minds. This requirement was narrowed in Bell Atlantic Corp. v. Twombly, 550 U.S. 544 (2007), where Supreme Court listed such plus factors as (1) the motive to conspire, (2) acts not taken in self-interest, and (3) evidence of a traditional conspiracy, all of which are factors that become evident on demonstration of intent.[16]
Interpretation of “Agreement” under the Competition Act, 2002
The Competition Act, 2002 seeks to interpret the term agreement mentioned in Section 2(b) to include any arrangement, understanding or act done in concert regardless of the oral, written, formal or law enforcement of the act. It is in this sense that this definition includes tacit and explicit collusion.
Presumption of Anti-Competitive Conduct and Emerging Challenges
The presence of a meeting of minds is a necessary component for an agreement which is supported by direct evidence. Moreover, the historic case of Builder association of India vs. The Competition Commission it was held that the cement manufacturers guilty of price-fixing (Cement Manufacturers Association 2012), which was based on co-ordinated price increases, as well as sharing of data, although there was no formal agreement detected.[17] Section 3(3) of the act provides that, horizontal agreements, including bid rigging, market allocation, output restriction, as well as price fixing are considered illegal per se. If in near future, AI relied upon by the different companies decided to engage in the process of tactic collusion, it will lead to prize fixing in the markets. Similar to what happened in the case of NRAI vs. Zomato & Swiggy and such agreement are therefore assumed to have an appreciable negative impact on competition as contained in Section 3(1).[18]
STRENGTHENING THE COMPETITION ACT FOR ALGORITHM-DRIVEN MARKETS
The existing regulatory policy of the Competition Commission of India (CCI) fails to provide the necessary solutions to the challenges posed by algorithmic collusion, especially when autonomous or self-learning algorithms are involved.[19] The existing statutory framework continues to follow an archaic paradigm of competition law that is only applied to express agreements or concerted practices among businesses. However, there may still be the risk of so-called hidden algorithmic collusion, without involving an active human decision-making process, which leaves a gap in the existing enforcement framework.[20]
Pre-Deployment Review and Algorithmic Audits
To solve this problem, some structural and regulatory changes can be considered. One should begin by amending the Competition Act, 2002, to clearly address anti-competitive behaviour based on algorithms. An explicit exception might be added that specifically addresses algorithmic pricing, market behaviour artificially controlled by artificial intelligence, and autonomous decision-making technologies. This provision would allow the CCI to develop custom enforcement tools and foster understanding of the law in situations where anti-competitive results arise from automated procedures rather than direct human dealings. The CCI could, under this suggested amendment, set up a specialised expert committee with the powers vested in it by the statute. Such a committee should be comprised of economists, data scientists, and experts in competition law who are technically knowledgeable about complicated algorithmic market behaviour. Their activity would be to evaluate the competitive implications of algorithmic systems, monitor risks emerging in digital markets, and provide independent expert evaluation to support the Commission in research and policy-making.
Second, a pre-deployment or pre-approval system ought to be implemented for algorithmic pricing and market algorithms that are highly risky.[21] It may also be mandatory that, before such algorithms are released to the market, companies place them under regulatory review to assess their potential to create anti-competitive effects. If the algorithm indicates a probability of promoting price collusion, elimination, or market inaccuracies, the certificate of approval may be refused. On the other hand, when the algorithm is not at risk of competition, it can also be allowed to enter the market, provided it is regularly monitored for compliance. An annual audit and market review by the suggested expert committee also confirms that the algorithms deployed are within the realm of competitiveness. Regulatory changes in the European Union support this practice. Compliance by design has become a more prominent focus of the EU, in which companies ought to ensure that their algorithmic systems are designed and implemented in a manner consistent with their duties under competition law.[22] Under the new regulations, such as the EU AI Regulation and Articles 101 and 102 of the Treaty on the Functioning of the European Union, companies will be required to track the consequences of their algorithms and rectify anti-competitive effects, even when unintended.[23]
Lessons from the European Union: The DMA and AI Act
Similarly, broader digital market laws, including the Digital Markets Act (DMA), have active obligations on big digital platforms to avoid anti-competitive or antitrust behaviour in algorithm-based markets.[24] The DMA will make digital markets competitive and fair through imposing accountability on dominant digital "gatekeepers" and avoiding market practices that can manipulate competition. Those developments illustrate how competition regulators can integrate both ex-ante regulatory mechanisms and classical antitrust enforcement.[25]
The history of cartel regulation demonstrates that the competition law has been historically constructed on the basis of the idea of agreements and the presence of the meeting of the minds between the market actors. The legal systems like the Sherman Antitrust Act and the Competition Act, 2002 have thus concerned themselves with determining the presence of coordinated intent and communication as a way of proving collusion. But with increasing reliance on automated and data-driven pricing systems in markets, these conventional assumptions might also prove insufficient, which leaves critical questions regarding the ways the competition law would deal with coordination that should emerge without humans expressing it.
1. Org. for Econ. Co-operation & Dev. (OECD), Algorithms and Collusion: Competition Policy in the Digital Age, DAF/COMP(2017)4 (2017), available at https://one.oecd.org/document/DAF/COMP(2017)4/en/pdf. ↑
2. Alejandro Guerrero Perez, Ombline Ancelin, David Trapp, Andrea Pomana & Sarah Smith, EU Competition Authorities Zero in on Antitrust Risks of Algorithmic Pricing, Global Competition Review: Digital Markets Guide (5th ed.) (Sept. 22, 2025), https://globalcompetitionreview.com/guide/digital-markets-guide/fifth-edition/article/eu-competition-authorities-zero-in-antitrust-risks-of-algorithmic-pricing. ↑
3. Anshika Gupta, From Cartels to Code: Can India's Digital Competition Law Tame AI Gatekeepers?, Cyber Blog India (Jan. 18, 2026), https://cyberblogindia.in/from-cartels-to-code-can-indias-digital-competition-law-tame-ai-gatekeepers/. ↑
4. Ayush Agrawal, Algorithmic Collusion and the Limits of Agreement under Indian Competition Law, Indian Review of Corporate and Commercial Laws Blog (Feb. 21, 2026), https://www.irccl.in/post/algorithmic-collusion-and-the-limits-of-agreement-under-indian-competition-law. ↑
5. Arpita Gupta, Algorithmic Collusion and Its Challenges to Antitrust Regulations, Int'l J. Legal Rsch. & Analysis, Vol. II, Issue 7, at 1 (Mar. 2025), https://www.ijlra.com/details/algorithmic-collusion-and-its-challenges-to-antitrust-regulations-by-arpita-gupta. ↑
6. Digvijay R. Singh, Algorithmic Collusion: Can the Competition Act Protect against Self-Learning Algorithms?, IndiaCorpLaw (Jan. 6, 2022), https://indiacorplaw.in/2022/01/06/algorithmic-collusion-can-the-competition-act-protect-against-self-learning-algorithms/. ↑
7. Aakriti Bansal, The Invisible Price Cartels: How AI Algorithms Could Be Colluding Without Human Intent, Medianama (Oct. 8, 2025), https://www.medianama.com/2025/10/223-ai-algorithms-price-collusion-human-intent-cci/. ↑
8. Ibid. ↑
9. Christophe Samuel Hutchinson, Gulnara Fliurovna Ruchkina & Sergei Guerasimovich Pavlikov, Tacit Collusion on Steroids: The Potential Risks for Competition Resulting from the Use of Algorithm Technology by Companies, Sustainability Vol. 13, No. 2, 951 (2021), https://www.mdpi.com/2071-1050/13/2/951. ↑
10. Org. for Econ. Co-operation & Dev. (OECD), Algorithms and Collusion: Competition Policy in the Digital Age, OECD Roundtables on Competition Policy Papers No. 206, OECD Publ'g, Paris (2017), https://www.oecd.org/content/dam/oecd/en/publications/reports/2017/05/algorithms-and-collusion-competition-policy-in-the-digital-age_02371a73/258dcb14-en.pdf. ↑
11. Erica Stanford, Autonomous AI: Who Is Responsible When AI Acts Autonomously and Things Go Wrong?, Global Legal Insights – AI, Machine Learning & Big Data Laws and Regulations (2025) (May 15, 2025), https://www.globallegalinsights.com/practice-areas/ai-machine-learning-and-big-data-laws-and-regulations/autonomous-ai-who-is-responsible-when-ai-acts-autonomously-and-things-go-wrong/. ↑
12. Ayush Agrawal, Algorithmic Collusion and the Limits of Agreement under Indian Competition Law, Indian Rev. Corp. & Com. Laws Blog (Feb. 21, 2026), https://www.irccl.in/post/algorithmic-collusion-and-the-limits-of-agreement-under-indian-competition-law. ↑
13. Saksham Agrawal, Addressing Algorithmic Collusion in Indian Competition Law, CBCL Blog (Dec. 16, 2025), https://cbcl.nliu.ac.in/competition-law/addressing-algorithmic-collusion-in-indian-competition-law/. ↑
14. Aarav Sharma, Algorithmic Cartels and Antitrust Law: How Pricing Algorithms Are Redefining Market Collusion, Gujarat J. L. & Ethics (GJLE) (Mar. 2, 2026), https://gjle.in/2026/03/02/algorithmic-cartels-and-antitrust-law-how-pricing-algorithms-are-redefining-market-collusion/. ↑
15. Pál Szilágyi & András Tóth, Historical Developments of Cartel Regulation (May 23, 2016), Versenytükör 2016 (upcoming English Special Edition), available at SSRN. ↑
16. Bell Atl. Corp. v. Twombly, 550 U.S. 544 (2007). ↑
17. Builders Ass'n of India v. Competition Comm'n of India, Appeal No. 79 of 2012, Competition Appellate Tribunal (Dec. 10, 2012). ↑
18. National Restaurant Association of India v. Zomato Ltd. & Swiggy, Case No. 16 of 2021, Competition Comm'n of India (Apr. 4, 2022). ↑
19. Competition Comm'n 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. ↑
20. Saksham Agrawal, Addressing Algorithmic Collusion in Indian Competition Law, Centre for Bus. & Com. Laws (CBCL) Blog, NLIU (Dec. 16, 2025), https://cbcl.nliu.ac.in/competition-law/addressing-algorithmic-collusion-in-indian-competition-law/. ↑
21. Arpita Gupta, Algorithmic Collusion and Its Challenges to Antitrust Regulations, Int'l J. Legal Rsch. & Analysis Vol. II, Issue 7 (Mar. 2025), https://www.ijlra.com/details/algorithmic-collusion-and-its-challenges-to-antitrust-regulations-by-arpita-gupta. ↑
22. Michal S. Gal & Daniel L. Rubinfeld, Algorithmic Pricing and Antitrust: A New Challenge for Competition Law, J. Competition L. & Econ. (2024), https://www.sciencedirect.com/science/article/pii/S0267364924000517. ↑
23. European Commission, Competition Law Treaty Articles, https://competition-policy.ec.europa.eu/antitrust-and-cartels/legislation/competition-law-treaty-articles_en. ↑
24. Eur. Comm'n, Digital Markets Act, https://digital-markets-act.ec.europa.eu/index_en. ↑
25. Tilman Harmeling, DMA Gatekeepers: Their Role and Impact under the Digital Markets Act, Usercentrics (Oct. 25, 2023), https://usercentrics.com/knowledge-hub/role-of-gatekeepers-under-digital-markets-act/. ↑
Views are personal.