CODE AS AGREEMENT: SHARED PRICING ENGINES, HUB-AND-SPOKE LIABILITY AND THE LIMITS OF SECTION 3 OF THE COMPETITION ACT, 2002

Author

Tanish Patel,Priyansh Shrivastava,Ashmit Chauhan

Abstract

 When rival firms license the same pricing engine, the question whether they have ‘agreed’ on price increasingly turns on facts that engineers, not lawyers, create: what data the engine ingests, whether it pools rivals’ non-public inputs, and whether it recommends prices or executes them. Since the Competition (Amendment) Act, 2023, section 3(3) of the Competition Act, 2002 has carried an express hub-and-spoke proviso that presumes a non-competing enterprise to be part of a horizontal agreement if it ‘participates or intends to participate’ in its furtherance. No reported Indian decision has yet applied that proviso to shared pricing software. This article argues that the proviso supplies the words but not the evidentiary or doctrinal tools to apply them to a software vendor’s design choices. Drawing on the Third Circuit’s July 2026 decision in Cornish-Adebiyi v Caesars Entertainment, the Ninth Circuit’s decision in Gibson v Cendyn Group, the RealPage consent judgment, California’s AB 325, the Court of Justice’s judgment in Eturas and the Competition Commission of India’s 2025 Market Study on Artificial Intelligence and Competition, it proposes a design-based test of the vendor’s intent, paired with a rebuttable safe harbour for tools that run only on a firm’s own or public data. It also argues that the problem of autonomous self-learning agents should be addressed through deployment-based responsibility and audit, not by diluting the agreement requirement.

Keywords

algorithmic collusion; hub-and-spoke cartels; Competition Act 2002; section 3(3); pricing software; information exchange; Competition Commission of India

Full Text:

Download Paper PDF

References

The question in the shared-engine cases is an engineering one: what the tool does with data, and what users and vendors know about it. The American and European authorities converge on that answer. Common use of a vendor is not collusion; the pooling of rivals’ confidential data in a tool that each rival knows its competitors use, combined with features that discourage independent pricing, can be.

India has the statutory language to ask the question. The 2023 proviso, read against its drafting history, makes a vendor’s intention to further a horizontal agreement sufficient to presume it part of that agreement. But the Act and the CCI’s practice do not yet say how that intention is to be shown for a software product, and the procedural framework (global-turnover penalties, the exclusion of section 3(3) cases from commitments, and uncertain leniency for hubs) raises the cost of getting the answer wrong. Without published guidance, forensic capacity and a safe harbour, the proviso risks being either unused or over-applied.

A design-based approach to intent offers a principled middle path. It reads the proviso’s words in light of their object, uses the engineering record that already exists, proves the rim and the hub’s role with the same evidence, and leaves genuinely independent algorithmic pricing to the audit track rather than stretching the concept of agreement past breaking point. It also gives engineers something lawyers rarely can: a clear account of which design choices carry legal risk, made before the code ships rather than after the CCI arrives.

Share your valuable work from Social Media Buttons