On May 20, 2026, the Calcutta High Court delivered what may become the most consequential AI ruling of the year in India. Justice Ravi Krishan Kapur dismissed IndiaMART InterMesh Ltd's interim relief application seeking to force OpenAI to surface IndiaMART links inside ChatGPT responses. The ruling, reported by MediaNama, reverses the same judge's December 2025 view that had found a "strong prima facie case" of selective discrimination. After hearing OpenAI, the court concluded there was no prima facie case and that the balance of convenience weighed against intervention.
For a country where ChatGPT now serves a reported 100 million weekly users — the second-largest user base in the world — this is the first judicial answer to a fundamental question: does an Indian business have a legal right to be linked, cited, or surfaced by a foreign-built large language model? The answer, at least at the interim stage, is no. That is the correct outcome, and it should shape how India approaches AI regulation going forward.
What the Court Got Right
IndiaMART's argument was, in essence, that ChatGPT's failure to recommend its B2B marketplace amounted to commercial discrimination. The December 2025 order had agreed, calling the exclusion illogical. But "illogical" is not the same as "unlawful." Generative AI systems are not neutral switchboards; they are probabilistic models whose outputs reflect training data, retrieval choices, ranking heuristics, and reinforcement learning from human feedback. A model that does not return a specific URL has not "refused service" — it has simply not been induced, by its inputs and weights, to produce that output.
Treating non-mention as actionable harm would turn every LLM into a compelled-carriage system. Each commercial platform left out of a response could file suit; each underrepresented brand could demand surfacing. The compliance cost of operating in India would balloon, and the inevitable response — visible already in how some AI vendors handle EU-only restrictions — would be to geo-fence or down-rank Indian outputs entirely. That outcome would be catastrophic for the millions of Indians who currently rely on these tools for work, study, and small-business research.
The "Platform Neutrality" Trap
Commentators including Navneet Sharma, VB Padode Chair Professor of Competition Policy at Vijaybhoomi University, have argued that AI assistants are now "actively constructing the consumer's informational universe," and that exclusion from their outputs "can materially affect market access." That framing has surface appeal. But it conflates two very different things: the editorial choices of a model, and the gatekeeping power of a regulated bottleneck.
Search engines and app stores have been treated as bottleneck infrastructures because of their market power and architectural role as singular discovery layers. The Indian generative AI market looks nothing like that. ChatGPT competes head-to-head with Google's Gemini, Anthropic's Claude, Meta AI inside WhatsApp, Perplexity, Microsoft Copilot, and a growing cohort of Indian-built models including Sarvam AI, Krutrim, and open-weights derivatives of Llama and Mistral. If IndiaMART is unhappy with one model's outputs, the open-weight ecosystem already offers the option to fine-tune or self-host an alternative. The remedy is competition, not compulsion.
Open Source Is the Answer, Not Mandated Linking
This is where India's policy posture matters. The IndiaAI Mission, approved by the Union Cabinet in March 2024, explicitly prioritises sovereign foundation models and open compute access. That bet — that India should grow its own AI capability rather than regulate foreign models into compliance — is the right one. A judicial precedent forcing OpenAI to surface specific Indian businesses would have signalled the opposite: that India will treat AI platforms as common carriers subject to ad hoc commercial demands. That posture chills both foreign investment and domestic experimentation, because open-source builders downstream would inherit the same liability exposure.
The Calcutta High Court's restraint is consistent with the spirit of Section 79 of the Information Technology Act, 2000, which provides safe harbour to intermediaries for third-party content. While LLM outputs are arguably first-party generated, the underlying logic — that platforms cannot be held liable for every editorial micro-decision in a system processing billions of queries — applies with equal force.
Real Risks, Different Remedies
None of this means generative AI raises no genuine policy concerns. The same week as the Calcutta ruling, Ars Technica reported on TeamPCP, a hacker group running a software supply chain attack spree against open-source repositories, including a confirmed GitHub breach via a poisoned VSCode extension. That is a real harm — the kind that warrants regulatory attention, transparency obligations, and security baselines for AI-assisted code generation.
The lesson is one of proportionality. India should regulate AI for things AI actually does badly: model security, deepfakes, election integrity, and downstream supply-chain poisoning. It should not regulate AI for failing to surface every commercial player who would prefer more visibility. Confusing the two will hand India's AI future to whichever foreign lobby pleads hardest in court.
What Comes Next
The Calcutta HC ruling is interim, and larger constitutional questions around AI accountability under the IT Act and the proposed Digital India Act framework remain open. The Supreme Court will eventually weigh in. When it does, it should anchor on the principle Justice Kapur reached: competitive harm requires more than non-mention, and the cost of forced surfacing falls not on OpenAI alone but on every Indian builder hoping to ship a model of their own.
For now, the ruling deserves cautious applause. India has chosen, on its first real test, the path that keeps the AI market open, contestable, and innovation-friendly. That is the correct foundation for everything that comes next.