Algorithmic Collusion 2.0: Why India’s Ex-Post Competition Framework Cannot Handle AI-Driven Market Coordination

I. Introduction

Across jurisdictions, algorithmic pricing is now routine in sectors like ride-hailing, airlines, and e-commerce. These systems continuously adjust prices based on demand, competitor behaviour and consumer data, and often “learn” to mirror rivals’ pricing. Over time, many adopt a tacit tit-for-tat pattern, discovering that coordinated price increases maximise profits. Markets may thus converge to persistently high prices without any explicit agreement or human intent. Even the CCI has warned that predictive AI can mask collusive outcomes under the guise of dynamic pricing.[1]

By contrast, the Indian Competition Act, 2002 is built on an ex-post, human-centric model.[2] Evidence of a prior agreement like emails, meetings, or other proof of a conscious “meeting of minds” is required.[3] This framework captures traditional cartels but not coordination produced by self-learning algorithms. The CCI’s own market study notes that AI systems can autonomously align strategic behaviour, including pricing, generating cartel-like outcomes with no human involvement. In effect, conventional law looks for a deliberative cartel, while algorithmic collusion is a faceless, coded one.

Recent Indian cases illustrate these tensions. In Samir Agrawal v. CCI (2023), the Supreme Court declined to treat Ola and Uber’s surge-pricing algorithms as illegal collusion, finding no evidence of any driver agreement. By contrast, in 2022 the CCI fined online travel agencies MakeMyTrip (GoIbibo) and budget hotel chain Oyo ₹3.92 billion for imposing broad “price-parity” and exclusivity clauses on hotels. And in the 2021 “Beer Cartel” case, CCI penalized major brewers for traditional price-fixing. These actions show regulators scrutinize digital platform practices, but the investigative tools remain firmly retrospective. India’s reliance on a purely ex-post, agreement-centric model risks leaving AI-enabled coordination largely beyond its reach.

II. How AI Enables “Collusion 2.0”

Algorithms can generate tacit collusion that operates very differently from human cartels. Modern AI systems do not simply follow preset rules; they learn from market feedback. Research shows that reinforcement-learning and other adaptive pricing algorithms can independently converge to supra-competitive prices through repeated interactions, even without any instruction to collude.

Three factors intensify this risk. First, algorithms monitor rivals in real time and quickly punish price cuts, stabilising collusive outcomes. Second, they process granular, high-frequency data—competitor prices, demand shifts, and inventory levels—allowing rapid alignment on profitable price points. Third, the opacity of complex AI models hides this coordination: the resulting uniform prices look like ordinary optimisation rather than cartel conduct. When competing firms use similar pricing engines, prices tend to ratchet upward, creating cartel-like effects that regulators struggle to detect because no communication—or evidence trail—exists.

III. Why India’s Ex-Post Framework Struggles

India’s antitrust regime (Section 3 of the Act) is geared to agreements and discernible collusion.[4] As a result, algorithmic coordination faces three core obstacles under current law:

• First, no traditional “agreement.” Autonomous pricing algorithms do not contract or communicate like humans. A code-driven optimization process produces aligned prices through computation, not consensus. India’s law requires an “agreement,” arrangement or understanding among enterprises. But an algorithmic cartel has no boardroom handshake or written agreement to capture. In other words, firms may align prices by design of their algorithms without ever exchanging a word, easily evading Section 3’s agreement requirement.[5]

• Second, evidentiary limits. Even if the CCI suspects algorithmic coordination, uncovering it is hard. Proving causation would require detailed access to an algorithm’s code, data inputs, and decision logs – information beyond the reach of current procedures. Unlike a human cartel (where a whistleblower or leaked memo might exist), here the “smoking gun” is a statistical pattern deep in proprietary software. The Competition Act contains no explicit audit powers or documentation obligations for algorithms, so essential digital evidence can remain hidden.

• Third, speed of collusion vs. slow enforcement. AI pricing changes can happen in minutes or hours. By the time a CCI investigation and appeal conclude – a process that often takes years – an algorithmic cartel will have long stabilized and embedded its effects in the market (e.g. eliminating new entrants or shifting consumer habits). The “ultra vires” lag means even winning a case may come too late to restore competition.

These frictions have played out in practice. In Samir Agrawal v. CCI, the informant alleged that Uber and Ola acted as a “hub” facilitating a cartel among drivers. The CCI and Supreme Court rejected this, treating the platforms as “single economic entities” whose algorithms set fares unilaterally, and noting that independent drivers never explicitly agreed to fix prices. Without proof of a conscious agreement among drivers, the court held there was no Section 3 violation.[6] The case thus underscores the mismatch: in an algorithmic market, antitrust claims hinge on detecting intent, but the system’s coordination can be entirely tacit.

IV. Case Studies and the Indian Market Context

Algorithmic pricing is already widespread in India. Large Indian airlines use automated yield-management systems that change fares by the minute. Nationwide cab platforms like Ola and Uber adjust fares via surge algorithms. Major online retailers and marketplaces (Flipkart, Amazon India and their sellers) routinely employ dynamic-pricing engines that react to rival discounts and inventory shifts. For example, Amazon’s algorithms adjust millions of product prices multiple times per day to stay competitive. In each sector, price updates happen too fast and complexly for consumers to easily detect.

Current enforcement, however, largely shoehorns these phenomena into legacy concepts. For instance, parallel price patterns in airlines are typically viewed as oligopolistic interdependence (a benign outcome) rather than collusion. Aggressive discounts on e-commerce platforms are often treated as potential resale price maintenance or abuse of dominance. But none of these doctrines truly captures a situation with no explicit cartel leader or agreement, yet algorithms are jointly lifting prices. As one commentator notes, algorithms may keep prices high “without even breaking the law” – regulators see price coordination but no culpable communication.

In the meantime, the CCI has also been active in seeking big cases in the digital markets with the conventional means. In October 2022 it fined MakeMyTrip/GoIbibo and Oyo on the ground of their featuring wide and exclusivity clauses in relation to the hotels. The case concerned a tangible deal (exclusive listing and deep discounts) as opposed to any computerized price program. Likewise, the current investigation of Amazon and Flipkart by CCI is focused on the alleged exclusive deals and predatory pricing techniques; in particular, in 2021, the Supreme Court of India permitted this inquiry to continue, indicating that the regulating authorities also focused on the behaviour of the online market. Another indicator of a typical cartel enforcement case is the 2021 decision of the Beer Cartel: major brewers were convicted of price coordination by CCI, which demonstrates that blatant collusion remains punishable even today. Those are examples of proactive policing in digital realms, depending on the uncovered apparent agreements or dominance misconducts. The current doctrine might be lacking in a statutory hook in the less blatant cases - where algorithms optimally adjust price adjustments concurrently.

V. The Case for an Ex-Ante Framework

In this light with such gaps, most analysts believe India must have an aggressive overlay to the competition policy. The mere intense enforcement will probably be too sluggish and limited. At the very least, regulating authorities may place transparency and audit requirements to the companies that implement advanced pricing algorithms in monopolistic markets. As an example, the companies may be made to self-audit their AI systems which entails recording of these fundamental parameters: what objective should the algorithm optimize (e.g. revenue maximization vs. market-share growth), what data does the algorithm utilize (competitor prices, demand metrics) and to which extent the algorithm is autonomous. These disclosures and disclosure of such kept confidentiality against other competitors would allow the antitrust authorities to signal suspicious coding issues in a timely manner. The new CCI AI report also suggests that companies ensure that there are mandatory competition-risk audits of their algorithms, which define the aims of algorithms, inputs of the algorithms and the rules that regulate decision-making.

There might also be compliance requirements that regulators may dictate with respect to the high risk pricing systems. An example is that the legislation may simply forbid the usage of a shared pricing software or even the algorithms that are aimed at a particular competitor. (California and other states of the US have started to prohibit some of the more popular pricing devices to anticipate collusion) Instead, there are proposals to add a rebuttable presumption of collusion in cases where the companies are fully aware of sharing the same algorithms and thus put the burden on the companies to demonstrate innocent motives. Most importantly, the Indian policy-makers are advised to enforce audit trails at the algorithmic level and That pricing records should be accessible to the competition authorities in real-time when examining digital industry. These ex-ante controls would not stay in place of the Competition Act, but instead they will supplement it with making algorithmic coordination more visible, and the anticompetitive designs will be thwarted prior to it happening.

VI. Foreign Models and Comparative Lessons

This is the direction that other jurisdictions are taking. In the EU, the 2022 Digital Markets Act (DMA) identifies the biggest named gatekeeper platforms and ex ante to be responsible to promote fair play.[7] Gatekeepers are also prohibited to self-preferencing (to favour their own products) and have to be interoperable and able to access data with their competitors - which is indirectly regulating the algorithmic ranking and pricing system. In fact, a recent EU analysis notes that the DMA (along with the new Digital Services and AI Acts) explicitly includes competition safeguards that “may curb the unfettered use of pricing algorithms”. Similarly, the UK’s new Digital Markets, Competition and Consumers Act (2024) grant the Competition and Markets Authority direct powers over dominant digital firms, including duties on pricing transparency and accuracy; the CMA has explicitly called for scrutinizing and testing AI-driven pricing models in fast-moving digital markets (e.g. requiring companies to explain how their pricing algorithms work).

In the United States, enforcers have also sharpened their focus on algorithmic pricing. The U.S. Federal Trade Commission (FTC) and Department of Justice have filed notable cases: for example, in September 2023 the FTC (with 19 states) sued Amazon, alleging that its secret algorithm “Project Nessie” raised Amazon’s prices to test competitor reactions and then locked in higher prices if rivals followed suit. A separate August 2024 DOJ lawsuit targeted RealPage, a property-management software firm, for facilitating collusion among landlords via its rental-pricing algorithm. U.S. lawmakers have also proposed bills (such as the Preventing Algorithmic Collusion Act) to prohibit competitors from using shared pricing software and to mandate transparency on pricing algorithms. The theme is unified throughout these regimes: The antitrust enforcement is currently being complemented with the rules that are designed to check or at least contain the algorithmic decision-making.

VII. Conclusion and Way Forward for India

The Indian economy is rapidly becoming AI-driven and digital-based, but the competition law still heavily presupposes human cartels in the country. The danger of no reform is that price coordination by AI could silently stabilize prices at a higher rate or monopolize a particular industry, which would go undetected by the previous approach. The antitrust institutions in India need to change. On the doctrinal level, the Indian courts and CCI must clearly acknowledge that collusion can also be based on the design of algorithms, and not on the verbal agreements, and adjust the understanding of meeting of minds to the new definition. In procedural terms, to invest more heavily in technical capacity, the CCI should build its own data-science resources and forensic software to study pricing algorithms as suggested by the recent policy research. On the regulatory domain, the Parliament is considered to embrace specialized transparency and audit regulations on digital markets as seen in the developed countries.

Overall, there should be a mixed solution: maintain the benefits of ex-post law (sanction against the evident illegal transactions) and implement ex-ante controls against the algorithmic collusion. This would assist India to evade the worst impact(s) of the Collusion 2.0 era. The CCI needs to boost its technical abilities and infrastructure as one recent report activates and mandates it to efficiently study and combat algorithm-based anti-competitiveness. It is only by modernizing its toolkit and its legislation that India will be able to make sure that the potential of AI-based markets does not turn into a threat to competition.

[1] Competition Commission of India, Market Study on Artificial Intelligence and Competition (October 2025).

[2] The Competition Act, 2002.

[3] The Competition Act, 2002, §3.

[4] The Competition Act, 2002, §3.

[5] Id.

[6] The Competition Act, 2002, §3.

[7] Regulation (EU) 2022/1925 of the European Parliament and of the Council of 14 September 2022 on contestable and fair markets in the digital sector (Digital Markets Act).