Anjnay Bansal & Dhanashri Dutta are students at Maharashtra National Law University, Mumbai
Introduction
It is surprising how liking one hilarious Ravi Kishan meme could flood your feed with hundreds more like it. This raises a central question: do social media platforms’ roles in recommending and distributing user-generated content make them a publisher, and if not, does their recommendation system nevertheless affect their entitlement to protection under Section 79 of the Information Technology Act, 2000 (IT Act)? The question gained urgency after the Central Government brought Meta’s recommendation system under its regulatory framework to determine the need to reclassify Meta’s status from a social media intermediary to a publisher.
How does this recommendation system work?
Before determining Meta’s liability as a publisher or intermediary, it’s worth understanding how this recommendation system works. Meta explains that multiple factors are at play, but mostly, the next video or reel or even a post a user gets fed is shaped by their interaction with the content they previously watched.
The complexity lies in the distinction between control over distribution and control over content. This distinction helps determine whether intermediaries’ algorithmic intervention amounts to editorial control sufficient enough to characterise it as a publisher, or whether it remains a technological mechanism through which an intermediary distributes UGC. The CDEI Online Targeting report, by the UK Government, notes that organic recommendations are driven by user interactions, while paid content is shaped by advertiser-defined targeting and the platform restricting itself to content optimisation. Furthermore, Meta ranks the content based on user interactions; this also includes lowering the ranks of content violating platform policy, yet the content remains created and uploaded by users. So, the question is not whether Meta exercises any control. The question is not whether Meta exercises any control. In our view, it does. The real question is whether that control extends to making editorial decisions. The answer, we argue, is no.
Is Meta a Publisher?
The IT (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 (IT Rules) define “Publisher of online curated content” as one that performs a “significant role in determining the online curated content being made available.” On its face, the definition remains too broad. However, upon reading it, one condition stands out: control over the content that is being made available to the users. A notification issued by the Ministry of Information & Broadcasting made the picture more vivid by clearly distinguishing the OTT Platforms as publishers of all online curated content from those intermediaries, which are based on user-generated content (UGC). Unlike OTT platforms, the videos, reels, or posts uploaded on these social media intermediaries like Meta are not created, commissioned, or selected by them as a part of an editorial catalogue. The recommendation system simply determines which existing pieces of content are likely to be relevant to the given users. Intermediaries hold the distribution control, but not the editorial control over the substance or creation of content.
In the case of Patterson v. Meta Platform, the New York Supreme Court placed reliance on the Force v. Facebook to hold that mere arrangement and display of third-party content through algorithms does not classify platforms like Facebook as ‘developer’ or ‘creator’ of that content.” The next question in the queue remains whether platforms using these recommendations still enjoy the safety net provided under Section 79 of the IT Act. The bench, further, distinguished Anderson v TikTok, noting that if recommendation algorithms converted third-party content as first-party content, it would expose the intermediaries like Instagram/Facebook/TikTok/Google/X to liability for every defamatory statement posted by a user. Therefore, on this point the picture remains fairly clear. However, the next question in the queue remains whether platforms using these recommendations still enjoy the safety net provided under Section 79 of the IT Act.
Do recommendations pierce the Section 79 Safety Net?
The protection to the intermediaries under Section 79 is not unknown. The nature of the protection being conditional requires certain thresholds under Section 79(2)(b) to be satisfied; this includes that the intermediary should not initiate the transmission, select the receiver of the transmission, or select or modify the information contained in the transmission.
The case of MySpace Inc. v. Super Cassettes Industries Ltd. provides a good starting point. The Delhi HC held that automated formatting and the inclusion of advertisements did not amount to selecting or modifying third-party content. Since anyone with internet access could view/upload the content, and the automated system did not choose the audience, the platform was not held liable under Section 79(2)(b).
A plausible point of distinction can be argued with respect to the human intervention that may be done by platforms for content violating the website policy. To respond, in the case of Shreya Singhal v. Union of India, the Supreme Court, while outlining the scope of the safety net, stated that the status of intermediary cannot be equated to complete absence of intervention.
Content moderation remains a prerequisite for fulfilling the due-diligence obligations under Rule 3 of the IT Rules, which operates alongside the statutory safe-harbour protection.
In Twitter v. Taamneh, the SCOTUS rejected the argument that the recommendation system plays an active role in hosting and recommending content uploaded by terrorist groups, holding that the platform built its recommendation model based on user inputs and history, which does not amount to the platform’s participation in or creation of the content.
Does this give Meta an invincible shield?
The position seems clear, neither does Meta stand as a publisher nor do the recommendations strip the intermediary of its safety net. But does that mean the platforms bear no liability over these algorithms? After all, though automated, the system is conceived and built by the platform itself. To address this limb, jurisdictions like the EU and the UK Safety Act offer a dynamic approach to catching up with AI’s fast pace.
The UK Safety Act recognises multiple forms of harm, and one of them remains: the manner of its dissemination. Section 10 imposes a duty on the service provider to regulate its functionalities, including algorithms, placing them under the purview of “Ofcom”, a regulatory body responsible for investigating services causing harm and enforcing the duties on such platforms. Where the service provider defaults, it exposes the service provider to the higher penalty of £18 million or 10% of global turnover. The EU Digital Services Act follows a similar logic, requiring “very large online platforms and search engines” to conduct risk assessments of their algorithmic system. Moreover, it imposes an obligation on the platforms to transparently declare the parameters over which the recommendation system works. The Commission can seek information for investigation and impose a penalty of up to 5% of average daily income or worldwide annual turnover per day penalties or a penalty of up to 6% of worldwide annual turnover upon infringement.
Rethinking Liability: Distributor, Not a Publisher
Reflecting back on Section 79, the exemption is not absolute either. Yet that does not imply mere control over the dissemination of UGC should make the intermediary liable for malice in the content itself. This is exactly where the need to distinguish liability for the subjectivity and the manner of dissemination of the content arises. To illustrate, for a video containing some defamatory content, the user who uploaded it should attract liability for the offence. Conversely, the intermediary’s liability should arise only where the concern is rooted in the platform’s own system, including a failure to assess or mitigate the reasonably foreseeable risk created by the manner in which the videos are ranked, amplified and distributed to users. In the righteous sense, therefore, Meta should retain its intermediary status while bearing an independent responsibility for recommendation system’s design and operation.
The approach finds its clear basis in the regulatory model emerging in the EU and UK which has demonstrated this line of separation, making these platforms accountable for their own lack of compliance where it results in systematic failures. For India, the first step involves MeitY should evolve the IT Rules to include specific obligations for platforms regarding these recommendation systems. Such a framework should require the platform to conduct periodic risk assessment and disclosure of the governing parameters along with operation of the ranking system. It should envisage further a regulatory role for MeitY, one that involves not only seeking information and subsequent investigation but also mandating corrective measures.
India should pivot from other jurisdictions in how it penalises non-complaince. To avoid the situation of conflict of interest that CCI faces- investigating and penalising, under one statute – and to avoid political forces meddling with the penalty decision, the apt course of action remains to allow MeitY or private individual to file a complaint before Adjudicating Officer appointed under Section 46 of the IT Act, with an appellate jurisdiction with the Telecom Disputes Settlement and Appellate Tribunal for ensuring penalties for persistent or serious violations.
Conclusion
Returning to where we began from, it becomes clear, Meta is not a Publisher and it’s recommendation system does not strip off intermediary safe harbor. However, the safety net does become a free pass to all possible liability. The better approach would be to distinguish between the user’s responsibility for the content, Meta’s responsibility for complying with intermediary obligations and independent responsibility for the design and operation of the recommendation system. Algorithmic intervention should not automatically be considered editorial authorship, just as intermediary status should not become regulatory immunity. The user remains the “chef” of the content, while Intermediaries act as the “waiter” deciding how it is served. The waiter need not become the chef but cannot disclaim accountability if the serving itself caused harm.