India’s intermediary liability framework has long struggled to reconcile copyright enforcement with the protections afforded to platforms under Section 79 of the IT Act. This tension has recently been amplified by the Delhi High Court’s directions seeking clarity from Meta on its policies governing access to its Rights Manager tool (a development we will examine in one of our forthcoming posts), as well as by the amendments to the IT Rules tightening intermediary obligations (discussed by Sonisha here). Against this backdrop, Anushka Aggarwal argues that the problem may lie deeper, not merely in how Section 79 is interpreted, but in whether it was ever the right statutory framework for addressing intermediary copyright liability in the first place. Anushka is a fifth-year student at the National Law School of India University, Bengaluru.
India’s ISP Copyright Problem – Self-Inflicted?
By Anushka Aggarwal
Every major digital economy has eventually been forced to answer the same question: when a user infringes copyright through an online platform, how much of that liability travels upstream to the platform itself? The United States answered it through a decade of litigation culminating, most recently, with Cox Communications v Sony Music. India has answered it, formally, through a combination of the Copyright Act and the IT Act, but the answer has never been coherent, and the pressure building around AI-generated content may soon make that impossible to ignore. India has never produced its equivalent of Cox Communications: No nine-figure damages verdict, no Supreme Court showdown over secondary copyright liability, no years of litigation exposing the doctrine’s internal contradictions. The conventional explanation is that Indian copyright holders seem to prefer website-blocking injunctions over damages claims, partly because the Copyright Act lacks a statutory damages multiplier like Section 504(c) of the Digital Millennium Copyright Act (US Copyright Law). The absence of high-stakes litigation has been read as doctrinal stability.
It isn’t. It is merely incoherence that has not been sufficiently questioned. The threshold question, i.e., when does an internet service provider become liable for its subscribers’ copyright infringement, and what level of knowledge triggers that liability, has never been coherently answered under Indian law. The reason is foundational – statutory misreading that courts have compounded rather than corrected, and which the 2026 amendments to India’s Intermediary Guidelines have now pushed to the breaking point.
The 2026 Rules Expose the Fault Line
Section 79 of the IT Act conditions its safe harbour on actual knowledge, and in MySpace Inc v Super Cassettes Industries Ltd (2016) the Delhi High Court held that an intermediary offering services beyond mere access (as most platforms now do) must satisfy Section 79(2)(b): it cannot initiate the transmission, select its receiver, or modify the content transmitted. For years this was a manageable constraint. The 2026 amendments to the IT (Intermediary Guidelines and Digital Media Ethics Code) Rules have made that complicated.
Rule 4(1A) introduces a knowledge standard for synthetic content under which a Significant Social Media Intermediary that “knowingly permitted” or “failed to act upon” violating material is deemed to have failed its due diligence obligations. The intent is to push platforms toward more active oversight of AI-generated content. The effect, under the Myspace reading, is to create a self-defeating compliance structure: Section 79(2)(b) of the IT Act conditions safe harbour on the intermediary not initiating, selecting, or modifying the content it transmits. Section 79(3)(b) provides that actual knowledge, the standard that forfeits safe harbour, arises when the intermediary knows of infringing content and fails to act. Rule 4(1A) explicitly encourages platforms to exercise editorial judgment over a category of content. However, that editorial judgment is precisely the kind of conduct that Section 79(2)(b) treats as disqualifying for safe harbour purposes. Platforms that comply with the Rules may thus, demonstrate the active role that strips them of the protection the Rules preserve.
This contradiction is not a drafting oversight that can be resolved by purposive interpretation. It is the structural consequence of subordinate legislation attempting to set a knowledge threshold whose parent statute was misread as the governing framework for civil copyright liability in the first place. The Rules are trying to do nuanced work, i.e., distinguishing the editorial choices that should attract scrutiny from the neutral hosting functions that should remain protected, through a statute never designed for that purpose.
The recent plea by Google and Meta before the Delhi High Court shows how the contradiction is not merely academic. They argued that forcing platforms to proactively detect and block unauthorised recordings of court proceedings would require abandoning the position Section 79(2)(b) obliges them to hold. Google’s argument was that it can act only on content specifically identified by URL once a court has adjudicated it unlawful, and anything more would require judging, ahead of any order, whether a video shows a court proceeding, whether the recording was authorised, and whether it breaks rules that vary from court to court, which is precisely the editorial judgment that forfeits safe harbour under the Shreya Singhal reading of actual knowledge. Meta made the same argument. The government’s own counsel has separately argued, in the unrelated X Corp litigation, that Shreya Singhal’s court-order threshold has become outdated. Thus, the tension Rule 4(1A) creates for AI-generated content is not confined to that rule, and any proactive-monitoring duty conflicts with the same structural feature of Section 79 that courts and platforms are actively litigating.
Where’s the Application?
The case law does not strictly demonstrate irreconcilable readings, but a failure to apply a correction in jurisprudence. In Kent RO Systems Ltd v Amit Kotak (Delhi HC, single judge, 2017), a registered design dispute, the court rejected the argument that an intermediary must proactively screen future listings once it has received a single complaint. On the statute’s text, this was correct since Section 79(3) requires “conspired, abetted, aided or induced,” standard(s) requiring common intention, and the Intermediary Guidelines impose a notice-and-takedown duty, not a standing surveillance requirement. The court noted that if Parliament had wanted an ongoing monitoring obligation, it would have said so. Christian Louboutin SAS v Nakul Bajaj (Delhi HC, single judge, 2018) reached the opposite conclusion in a trademark dispute, holding that a luxury marketplace’s active involvement in curating content moved it outside Section 79 altogether, and constructing a multi-factor active-passive test with no anchor in Section 79(2)(b)’s actual language. That test did not survive judicial scrutiny. The Division Bench in Amazon Seller Services v Amway India Enterprises (Delhi HC, division bench, 2020) noted that a Division Bench had already disapproved the Louboutin approach in Clues Network Pvt Ltd v L’Oréal (2019), and it adopted that correction, i.e., Section 79 draws no distinction between active and passive intermediaries, and the only question Section 79(2)(b) actually asks is whether the intermediary initiates the transmission, selects the receiver, or modifies the content. Grounded in the statute’s plain text and in MySpace’s own holding that Section 79 is an affirmative defence rather than an enforcement mechanism, Amazon v Amway is the better-reasoned decision, and it is a Division Bench correcting a single judge, not a coordinate split.
The doctrine’s actual failure shows up sometime later. IndiaMART Intermesh Ltd v PUMA SE (Delhi HC, division bench, 2025) applied Amazon v Amway’s test and restored safe harbour, but without engaging the one fact pattern that test was arguably built to catch: IndiaMART’s revenue model rewards premium sellers (some plausibly counterfeiters) with enhanced visibility which is a platform-driven promotion that the court’s inquiry should have required. The problem is not that Indian courts cannot agree on a standard. A Division Bench gave one, correctly, in 2020. The problem is that the correct standard is not being applied to the fact pattern it was designed for. The US principle from Ellison v Robertson, that the essential inquiry is the causal relationship between the platform’s financial benefit and infringing activity, would have required examining whether IndiaMART’s revenue architecture gave it reasonable grounds to believe infringement was occurring. That is the question Section 51(a)(ii) would have prompted (and which I contend is the correct approach). It is not the question Section 79 asks.
This is not a problem unique to India, and the US experience shows why the active/passive distinction, however courts draw it, eventually breaks down. The US experience makes clear that this is structural rather than incidental. Cox Communications exposed how the knowledge-plus-material-contribution test produces incoherence when applied uniformly across defendants with fundamentally different relationships to infringing content. Within weeks of the decision, a developer of an AI platform invoked it to argue that deliberately excluding copyright from content-filtering infrastructure, while actively filtering violence and pornography, constituted mere inaction insufficient to ground contributory liability. The argument exposed precisely the gap the framework leaves open: a defendant making affirmative engineering choices about which categories of harm to police is not a passive conduit, but a doctrine built around the knowledge-plus-contribution test has no way to say so. Justice Sotomayor’s concurrence in Cox Communications identified this directly, warning that the majority foreclosed a third theory grounded in aiding-and-abetting principles that would have captured exactly this conduct.
India’s active/passive inconsistency across Christian Louboutin, Amazon v Amway, and IndiaMART is not a product of poor judging. It is what any doctrine produces when it applies a single knowledge standard to defendants whose structural relationship to infringing content differs fundamentally, and when the knowledge standard being applied was borrowed from a statute that was never designed for the purpose.
Thus, the question India’s doctrine has never coherently answered is not whether Section 79 applies. It is whether Section 79 was ever the right place to look.
