Cartels Without Conspiracy: Regulating Ai-Driven Algorithmic Collusion Under The Competition Act, 2002

Abstract

Competition law has always presumed that cartels require a meeting of minds. Section 2(b) and Section 3(3) of the Competition Act, 2002 define an anti-competitive ‘agreement’ in terms broad enough to capture tacit understandings, but they still anchor liability to human coordination. Pricing algorithms deployed across e-commerce, travel, ride-hailing, and fuel retail in India now routinely converge on supra-competitive prices without any exchange of intent between rival firms. This paper maps four archetypes of algorithmic coordination messenger, hub-and-spoke, predictable-agent, and autonomous ‘digital eye’ collusion against the doctrinal architecture of the Competition Act, 2002, and asks whether the 2023 amendment’s codification of hub-and-spoke liability closes the gap or merely narrows it. Drawing on the Competition Commission of India’s treatment of the domestic airlines pricing-software matter, the Uber hub-and-spoke proceedings, comparative developments including Lithuania’s Eturas ruling, French and European regulatory proposals for algorithmic audits, and the empirical economics literature on reinforcement-learning collusion, the paper argues that the agreement paradigm cannot, without reform, reach autonomous algorithmic coordination. It proposes a structural-plus-effects framework for a new Section 3A, calibrated disclosure and audit obligations for high-risk pricing algorithms, a rebuttable presumption keyed to market structure and outcome, and safe harbours for verified compliance-by-design, alongside institutional coordination between the CCI and sectoral regulators.