The Cats are Gone. The Question Remains.

Somewhere between a dramatic affair and an axe fight starring human-like cats, 4.21 million people decided that AI-generated cat videos were peak content. No cameras, no people, nothing resembling human effort. Just an algorithm producing feline soap operas on a loop, raking in money, and becoming one of YouTube’s fastest-growing channels. Super Cat League was not alone. Sixteen channels, just like it, collectively sitting on 4.7 billion views and 10 million dollars in annual earnings, were doing the exact same thing until YouTube began deleting all of it. The reason was that it fell into the category of “AI Slop,” meaning low-quality, mass-produced content that was surreal, uncanny, or grotesque. 

In the same cleanup, Screen Culture, an India-based channel with over a billion views, was also terminated. Its creators had labeled their AI-generated fake movie trailers as “parody,” “fan trailer,” and “concept trailer” but the label did not save them—because the system read the pattern, not the label. 

Unlike YouTube’s enforcement of its terms of service, what follows is a state-imposed statutory obligation, backed by the threat of criminal liability and safe harbour loss, that operates on the same logic but with constitutional consequences that private platform decisions do not carry.

India now gives the filter significantly more power and less time to think. The Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026, (hereinafter referred to as ‘IT Rules, 2026’) operative from February 20, 2026 introduced a regulatory framework for “synthetically generated information” or SGI.

Existing literature on the Rules, raises related concerns independently. This piece stands apart by placing satire, rather than automated censorship at large, at the centre of the analysis thereby examining three problems. First, it highlights the strike at satire because it represents the most constitutionally vulnerable form of expression under the new Rules. Satire requires false depiction to function, yet the Rules treat all false depictions as inherently unlawful; and the only space carved out for satire exists in a document the government itself says has no legal force. Second, it examines whether the government’s power in the 2026 Rules goes against the Bombay High Court’s ruling in Kunal Kamra v. Union of India. Third, it questions whether the automated systems on which the Rules depend are capable of performing the tasks assigned to them, and what their implications on Fundamental Rights are when they fail.

Just Label it They Said… 

A significant issue regarding the IT Rules, 2026, is the absence of any explicit exception for artistic and literary expressions within the definition of SGI. The exclusions listed under Rule 2(1)(wa) are narrowly confined to routine, good-faith editing or creation of documents/educational materials and the use of tools for accessibility, clarity or translation while keeping satire, parody, review and criticism outside the scope of legal protection.

Although the Frequently Asked Questions (FAQs) document released by MeitY highlights that “lawful uses such as labelled satire or creative synthetic works” may be permitted, this lacks the force of law as the FAQ itself admits that it is not a legal document and does not “replace, amend or alter” the IT Rules. In K.P. Varghese v. Income Tax Officer, the Supreme Court affirmed that administrative guidance cannot override statutory language when direct conflict exists.  There is an inherent conflict between the black-letter law of the Rules and the administrative guidance in the FAQs as under Rule 3(3)(a)(i)which say intermediaries must deploy technical measures to ensure users cannot “create, generate, modify, or alter” SGI that violates any law. Specifically, Rule 3(3)(a)(i)(IV) prohibits the creation of SGI that “falsely depicts or portrays a natural person or real-world event by misrepresenting… in a manner that is likely to deceive”. The central question is: how can one create satire or parody in the first place? Satire, by its very nature, often involves a false depiction or misrepresentation of a person’s conduct or statements for comedic or critical effect. If the SGI is realistic enough to meet the definition of being indistinguishable from a natural person, it will almost certainly be likely to deceive a viewer at first glance. 

In the Salam Samarjeet Singh v. High Court of Manipur, the Supreme Court emphasised strict adherence to statutory specifics over administrative clarifications. In the present scenario, although FAQ 7 suggests that such content may be permitted if labelled, the underlying Rule 3(3)(a)(i) prevents the very act of creation if the content is deemed deceptive. Therefore, an FAQ cannot bridge a legal vacuum that contradicts the primary text and such construction suggests that a creator is prohibited from generating parody-like content regardless of their good faith intent or subsequent labelling. Additionally, such an omission stands in stark contrast to international standards like Article 85 of the GDPR, which mandates that Member States must reconcile data protection with the right to freedom of expression and information, specifically requiring exemptions for processing carried out for journalistic, academic, artistic, or literary expression. Therefore, such exclusion leaves creators vulnerable to shifting administrative interpretations

Moreover, one of the major issues is the non-differentiation between harmful and harmless content. Rule 3(3)(a)(i)(IV) is triggered on the basis of a “tendency to deceive,” however, what the legislators have failed to account for is that deception does not always equate to injury. Examples of the same are AI-generated fan trailers as highlighted earlier, which may deceive a viewer into thinking a new film is coming, yet they cause no reputational or financial harm to the actors involved. Similarly, hyper-realistic videos of politicians singing or acting in films circulating near elections may be indistinguishable from a real person but viewers typically understand them as satire. However, under the current Rules, such satirical or political deepfakes are lumped together with malicious deepfakes because they meet the broad threshold of “misrepresentation likely to deceive”.

This lack of differentiation is further increased by Rule 3(2)(a) which allows any user to be a complainant thereby inviting bad-faith reporting to silence parody or social commentary. International jurisprudence highlights the dangers of such broad framing as reflected in Kohls v. Bonta, wherein US District Court struck down overly vague standards, requiring restrictions only for deepfakes causing concrete harm like voter coercion, not mere deception. Similarly, Minnesota narrowly tailored the restriction to content created with the intent to injure a candidate or influence an election without consent. However, such thresholds are completely overlooked in the IT Rules, 2026.

The IT Rules, 2026 fail the proportionality test established in Modern Dental College v. State of Madhya Pradesh and reaffirmed in Puttaswamy II which mandates that any restriction on Article 19(1)(a) must satisfy four distinct prongs i.e., legitimacy, suitability, necessity and a balance of interests. While the objective of shielding citizens from the unique challenges of SGI is legitimate, the Rules fail the remaining prongs. On suitability, a blanket ban on creating any  SGI that “falsely depicts” a person in a manner “likely to deceive” doesn’t proportionately target only harmful cases. On necessity, lesser alternatives exist like notice and takedown regimes that work internationally. On balance, the Rules overwhelmingly privilege state interest in preventing deception without accounting for legitimate artistic expression, thereby failing to balance Fundamental Rights with regulatory objectives. By treating all realistic synthetic portrayals as inherently unlawful and imposing a three-hour takedown window, the Rules create collateral censorship forcing intermediaries into automated self-censorship to preserve Safe Harbour.

Same Problem, New Rules: Kunal Kamra v. UoI Round 2? 

Another major issue with respect to the IT Rules, 2026, is the legal parity established between a court order and a reasoned intimation from a government appointed Authorised officer. Under Rule 3(1)(d)(ii)(I), an intimation issued by an officer not below the rank of Joint Secretary to the Government of India or to the State Government is treated as “actual knowledge” which triggers a mandatory obligation on the intermediaries to remove the content within a strict three-hour window. This regulatory introduction is problematic because it empowers the Executive to unilaterally adjudicate whether the content violates the broad and subjective categories namely public order, morality, decency or defamation. 

This resurrects the core constitutional infirmities identified by the Bombay High Court in Kunal Kamra v. Union of India. The proceedings observed that a government-controlled entity cannot be the final authority on what constitutes fake or false information regarding the business of the Central Government as such mechanism results in an unilateral determination by the Executive itself. While Kunal Kamra addressed a government FCU for traditional content and the IT Rules address the issue of SGI, the procedural problem in both situations is identical i.e., the executive branch assuming the role of identifying unlawful content without prior judicial intervention. Since, under the Rules an executive officer i.e., Joint Secretary or Director to the Government of India or to the State Government issues an intimation regarding content they deem for example, defamatory, to the administration, the State, herein, functions simultaneously as the complainant and the judge thereby violating the principles of natural justice highlighted by the Supreme Court in A.K. Kraipak v. Union of India. Further, the three-hour period provided to the intermediaries to take down such content upon actual knowledge exemplifies such a situation as this compressed window creates a severe chilling effect, as intermediaries have no practical opportunity to legally verify the “reasoned intimation” or seek judicial intervention. 

Although, it may be argued that the Rules attempt to offer a safeguard via a periodic review by an officer not below the rank of Secretary, conducted once a month to ensure that intimation are necessary and proportionate, this is an internal executive audit rather than an independent check thereby not equating to a judicial led oversight. This review is conducted within the same branch of government and cannot provide a neutral check.

Three Hours to Kill a Joke  

Rule 3(3)(a)(i) requires intermediaries to deploy automated tools to prevent unlawful SGI from being created or shared. This creates a significant burden: platforms must build pre-moderation systems at their own cost, remove flagged content within three hours of a government intimation with no discretion to verify or delay, and accept safe harbour loss under Section 79 if the deadline is missed. 

This burden directly engages Article 19(1)(g) of the Constitution, the right to carry on any trade or profession. Under Article 19(6), restrictions on this right are constitutional only if they are reasonable and in the general public interest. This restriction fails that test as it imposes a blanket obligation on all pre-moderation rather than adopting a less restrictive alternative. A notice- and- action model, where platforms remove only specific content once notified of unlawfulness would achieve the same regulatory goal without requiring constant surveillance at platforms’ expense

This approach already works internationally. The EU’s Digital Services Act, Article 8 explicitly states that “no general obligation to monitor the information which providers of intermediary services transmit or store… shall be imposed on those providers.” As the Supreme Court held in Chintaman Rao v. State of Madhya Pradesh, a restriction that goes further than necessary to achieve a legitimate public interest cannot survive constitutional scrutiny. The same principle applies here. 

This concern is not new to Indian law. In Shreya Singhal v. Union of India, the Supreme Court read down Section 79(3)(b) of the IT Act, which is the provision that requires intermediaries to remove content upon receiving ‘actual knowledge’ of its unlawfulness, to hold that such knowledge must come only through a court order or a formal government direction issued under due process. Without this safeguard, the Court reasoned, intermediaries would be forced to make their own subjective judgments about whether content is unlawful, creating precisely the chilling effect that free speech protections are designed to prevent. As mentioned earlier, a Joint Secretary’s ‘reasoned intimation’ as actual knowledge and attaching a three-hour removal window that Shreya Singhal was designed to foreclose, except now with a significantly shorter deadline and no meaningful opportunity for the intermediary to seek independent legal or judicial verification before complying.

The Congressional research documents how automated systems remove content incorrectly: Facebook once flagged Walla Walla onion photos as sexually explicit, removing legitimate business ads. If vegetables cannot survive automated moderation, satirical deepfakes will most definitely not.

The bias concern runs deeper than general over-removal. Research shows bias in training data directly transfers to moderation systems; they repeat past errors rather than improving on them. A 2025 article confirms AI models inherit historical biases and struggle with non-standard formats like satire and cultural allusions. This is particularly concerning for India since regional language expression and culturally specific humour are precisely the content most likely to be misread. AI hiring tools, built on the same machine learning framework, have already demonstrated this in practice: in Mobley v. Workday, a US federal court certified class actions proving AI tools discriminate on race, age, and disability grounds. The IT Rules, 2026 require no audit, no explanation, and no correction mechanism when the same thing happens in content moderation.

The European Commission’s High-Level Expert Group on AI has set out seven requirements for trustworthy AI, covering human oversight, technical robustness, privacy, transparency, fairness, societal wellbeing, and accountability. The IT Rules, 2026 fall short on almost every count. Intermediaries were given 10 days to implement compliant systems, making meaningful human oversight integration impossible. There is no requirement to explain to a creator why their content was removed, which means appeals are essentially blind. When the window to act is just three hours and the law offers no protection for satire, platforms will always choose removal over risk.

Conclusion 

The authors acknowledge that the IT Rules, 2026, are a necessary response to grapple with the challenges posed by SGI, including deepfakes, misinformation and impersonation. However, such regulatory ambition must not overwhelm constitutional safeguards. The current framework risks creating a chilling effect on freedom of speech and expression under Article 19(1)(a) by subjecting forms of expression such as satire, parody, political commentary to broad regulations. 

Additionally, subjecting intermediaries to proactive monitoring obligations and shortening the takedown window, leading to a considerable narrowing of safe harbour under Section 79 of the IT Act. This not only violates Article 19(1)(g) through imposition of unreasonable restrictions but also impacts the ease of doing business in the country as well as intermediary autonomy. The reliance on automated tools compounds this further, as such systems are prone to error, bias and over-removal with no accountability measures. Therefore, the Legislature must incorporate safeguards such as intent to injure or actual harmful impact as threshold requirements.

The authors are 3rd Year B.A LLB(Hons.) Students at the National University of Advanced Legal Studies (NUALS), Kochi.