AI And Antitrust: Is India’s Competition Law Ready For Algorithmic Market Power?

Algorithms can now influence what consumers pay, which products they see and what offers they receive. But when these systems shape market outcomes, distinguishing aggressive competition from anti-competitive conduct can become far more difficult.

The Competition Commission of India (CCI), in its October 2025 Market Study on Artificial Intelligence and Competition, flagged risks including algorithmic collusion, personalised pricing, self-preferencing and barriers to entry. The question is whether the Competition Act, 2002, can adequately address these emerging forms of market power when the processes driving them are often opaque and responsibility for their outcomes is difficult to establish.

Competition law experts Parthasarathi Jha, and Aparna Mehra, broadly agree that the Competition Act, 2002, provides a framework for addressing several AI-related concerns. However, the experts also identify significant challenges in applying established legal tests when the conduct in question is generated or influenced by increasingly autonomous systems.

Section 3 of the Competition Act prohibits anti-competitive agreements. The difficulty arises when competing businesses use algorithms that learn from market conditions and potentially align their pricing without any direct communication between them.

As Mehra points out, the issue is not entirely new to Indian competition law. In Samir Agrawal v Competition Commission of India, the Supreme Court examined allegations that Ola and Uber's algorithmic pricing restricted the ability of drivers to compete on fares. Court declined to infer a cartel from platform-determined algorithmic pricing in the absence of evidence of horizontal coordination among drivers.

However, the CCI has yet to see a case of algorithmic collusion, where no human intervention is involved.

Mehra says that the existing framework can address many forms of algorithmic coordination where human decision-making or intervention is involved. Section 2(b) defines an agreement broadly enough to include arrangements and understandings, while Section 3 addresses anti-competitive agreements.

"The more difficult scenario is where self-learning or reinforcement-learning systems independently converge on supra-competitive prices without any human communication or coordination," she explains.

Mehra adds: "Indian competition law, like most jurisdictions, does not treat conscious parallelism alone as sufficient evidence of collusion."

The distinction matters because similar pricing, by itself, does not establish that competitors have agreed to fix prices. Where algorithms independently arrive at similar outcomes, the regulator would still need to establish the elements of a contravention under the Act.

Jha similarly considers the position unsettled. "The extent to which purely autonomous algorithmic coordination, absent any agreement or concerted action between enterprises, would fall within the provision therefore remains an open legal question," he says.

A second concern arises when a company uses AI to favour its own products or services over competing offerings.

Section 4 prohibits the abuse of a dominant position. The provision addresses, among other things, unfair or discriminatory conditions, denial of market access and the leveraging of dominance in one market to enter or protect another. The legal question is whether these established principles can adequately address preferential treatment generated through AI systems, particularly when the system's decision-making process is not transparent.

Jha believes the existing framework is sufficiently broad to examine such conduct. "The CCI has already dealt with preferential treatment in cases involving digital products and its AI market study also recognises the potential for algorithms to facilitate exclusionary self-preferencing," he mentions.

However, he cautions: "self-preferencing should not be regarded as abusive per se".

"Any assessment must be fact-specific, considering the nature of the conduct, its implications for competition and market access, and legitimate business justifications, including efficiencies and consumer benefits," Jha says.

Mehra also considers Section 4 capable of addressing AI-driven self-preferencing, provided the conduct meets the legal requirements for abuse of dominance.

She identifies the principal difficulty as evidentiary. AI systems may generate different outputs depending on the user, prompt or context, making it difficult to establish whether a competitor has been systematically disadvantaged.

"The principal difficulty is evidential," Mehra says.

She adds that enforcement may depend on technical audits, access to the system's ranking logic and systematic testing of its outputs.

The same data-driven systems can also shape what consumers pay. By estimating individual willingness to pay, AI can enable targeted discounts but may also facilitate exploitative or exclusionary pricing when used by a dominant firm.

Both experts emphasise that personalised pricing is not inherently anti-competitive.

Jha says: "Competition concerns may arise where such practices facilitate collusion or, in the case of dominant enterprises, result in unfair or discriminatory pricing or exclusionary conduct."

Under Section 4, the relevant inquiry would therefore depend on the company's market position, the nature of the pricing practice and its competitive effects. "The mere use of AI to differentiate prices should not, by itself, attract antitrust liability," Jha emphasises.

Elaborating the issue, Mehra draws a distinction between dynamic pricing, which responds to changing market conditions, and personalised pricing, which uses information about an individual buyer to determine the price offered to that person.

The distinction becomes important when AI enables businesses to estimate consumers' willingness to pay with greater precision. Such pricing may benefit consumers in some circumstances, but it could also raise concerns if a dominant company uses it to exploit customers, selectively target competitors' customers or exclude rivals, Mehra points out, referring to the Organisation for Economic Co-operation and Development (OECD) 2019 report Artificial Intelligence in Society.

However, most such concerns would only attract scrutiny under Section 4 where the firm is dominant, Mehra emphasises.

Despite these difficulties, neither expert considers the absence of AI-specific provisions, by itself, proof that India's competition law is inadequate.

Jha argues that the Act is technology-neutral and provides a sufficiently flexible framework to address anti-competitive conduct in AI-driven markets. He sees the immediate challenge as developing the technical and institutional capacity required to assess sophisticated AI systems and their competitive implications.

"The priority should therefore be effective, evidence-based enforcement of the existing framework, while preserving incentives for innovation and investment," he says.

Mehra, meanwhile, points to the difficulty of understanding how some AI systems arrive at particular outcomes. "Many AI systems operate as 'black boxes', making it challenging for the CCI or even the companies using them to understand how outcomes were generated", she highlights.

Where a system's internal decision-making is opaque, identifying whether competitors were favoured, prices were coordinated or exclusionary conduct occurred can become more difficult. "As a result, detection, attribution, and evidentiary challenges are likely to be key hurdles for both companies and competition authorities," Mehra says.

The central challenge for Indian competition law, therefore, may not be an AI-specific overhaul, but its adequacy will depend on whether the CCI can establish liability in cases where the economic effect is visible, but the conduct behind it is not.