REGULATING SYNTHETIC CONTENT IN INDIA: A CRITICAL ANALYSIS OF THE INFORMATION TECHNOLOGY (INTERMEDIARY GUIDELINES AND DIGITAL

Table of Contents

Introduction: Rise of Synthetic Content Regulation in India

In early 2026, a video can look completely real and a voice can sound exactly like your favourite politician, yet neither has any link to reality. That loss of trust is now the heart of India’s digital problem. Against this backdrop, the February 2026 amendments to the Information Technology Rules arrived with unusual speed and ambition.^[1] They did not wait for harm to occur. Instead, they tried to mark content at the moment of its creation.

This marks a shift from reactive enforcement to pre-emptive identification at the point of creation.

Shift from Reactive Moderation to Proactive Content Identification under IT Rules 2026

The February 2026 amendments, issued under Section 87 of the Information Technology Act, 2000,^[2] reshape intermediary liability. Before, platforms only had to act when someone reported unlawful content. This approach, from Shreya Singhal v. Union of India,^[3] kept intermediaries from having to constantly monitor everything.

The new framework does something different. It focuses on the origin of content rather than its effect. By introducing SGI within the due diligence obligations under Rules 3 and 4,^[4] the law now expects intermediaries to identify synthetic content at the stage of upload itself.

This creates a kind of ‘digital chain of custody,’ where content stays traceable from creation to circulation, not just flagged after harm spreads. The emphasis is no longer only on removing harmful material. It is on ensuring that synthetic material never appears as real in the first place.

At one level, this appears necessary. Deepfakes and voice clones spread faster than any legal notice can catch them. But at another level, this shift quietly changes the role of platforms. They are no longer passive carriers. They begin to resemble verification authorities.

Defining Synthetic Information Through Perception-Based Standards

A key feature of the amendments is the definition of SGI under Rule 2(1)(wa).^[5] The definition does not rely only on how content is created. It depends on how it is perceived. If a piece of audio or video appears real or authentic to a reasonable person, it falls within SGI.^[6]

That focus on perception rather than just the tool is smart. Harm comes from what fools the eye and ear, not from the code itself. At the same time, the law deliberately limits its scope. Text-only content is excluded. Routine edits such as colour correction or formatting are also kept outside.^[7] This shows a certain caution. The State’s focus is on content with strong visual or emotional impact. However, the boundary between parody and deception remains unclear.

Mandatory Labelling and Transparency Obligations for SGI

Perhaps the most visible part of the amendments is the requirement of labelling. Any permissible SGI must carry a clear and prominent disclosure. For videos, this may appear as an on-screen label. For audio, it must be announced at the beginning.^[8]

The goal is straightforward — tell users it is fake so they do not get fooled. The execution, though, gets messy fast. The earlier draft rules had fixed numerical standards, such as a minimum size for watermarks. The final version moves away from rigid percentages and adopts a qualitative standard.^[9]

This change gives flexibility to platforms. It allows them to adapt disclosures to different screen sizes and formats. At the same time, it creates a grey area. What counts as “prominent” may vary from one platform to another.

Beyond visible labels, the rules also introduce metadata embedding. Each piece of SGI must carry a form of hidden identification that traces it back to the Intermediary.^[10]

When the Label Stays Even After You Delete It
There is a detail worth pausing on. The whole point of embedding metadata into synthetic content is that it travels with the file. You can strip the visible label. You can re-upload it somewhere else. But the metadata stays. That is the design. It is meant to solve the problem of content jumping across platforms after someone tries to scrub it clean. Whether that actually works in practice is a separate question — but it is the intended design, even if its real-world performance is still uncertain.

Institutional Burden and Compliance Pressure on Intermediaries

The amendments do not sit in a vacuum. They bump directly into Article 19(1)(a) and the right to privacy that the Supreme Court recognised in Justice K.S. Puttaswamy v. Union of India.^[11] That makes proportionality a live issue, not a theoretical one.

But in practice, it forces creators to say something in a format the State decides about their own work. This becomes more difficult when satire or political parody enters the picture. If a satirical video must always carry an AI-generated label, does the label change how people read it? And if the label shapes interpretation, is the State doing something more than just informing viewers?

The proportionality test from Modern Dental College v. State of Madhya Pradesh^[12] asks whether a restriction is necessary and matched to the harm it targets. Transparency is a legitimate aim — no argument there. But applying one uniform standard to all synthetic or generated content — regardless of context, intent, or actual risk — might be stretching that aim too far. A harmless edited clip and a malicious deepfake sit inside the same framework. That feels like it needs more nuance.

Metadata embedding also raises a privacy concern that is easy to underestimate. If every piece of synthetic content carries a traceable identifier, anonymity gets harder to protect. This is not just about bad actors. Whistleblowers, political critics, and ordinary users who do not want their creative work tracked back to them are all affected by the same rule.^[13]

Erosion of Safe Harbour and Shift to Proactive Liability

For a long time, platforms in India could rely on Section 79 of the IT Act for protection.^[14] The basic deal was: act on notice, do not monitor proactively, and you are shielded from liability. Shreya Singhal v. Union of India^[15] backed this up clearly. Courts said general monitoring obligations were too broad a burden to place on intermediaries.

The February 2026 amendments quietly move away from that. Now platforms — especially Significant Social Media Intermediaries — must detect and label synthetic content before harm occurs. Before. Not after receiving a complaint, not after a court order. Before. For SSMIs, this even includes verifying what users themselves declare about their content, using technical tools.^[16]

That is a meaningful change. The burden has shifted from reactive to proactive. Whether platforms can actually carry that burden accurately, consistently, and without chilling legitimate expression is still an open question.

Conclusion: Balancing Transparency, Truth, and Free Expression

The February 2026 changes try to deal with deepfakes, which can harm someone’s reputation very fast. But the problem is, they depend a lot on detection tools that are not always reliable and strict time limits that may push platforms to remove too much content.^[17] A better way could be clear protection for honest parody, more time for smaller platforms to check content properly, and regular checks on how labelling systems work. And in the end, the real issue stays the same — can the law protect truth without shrinking free speech?^[18]

Footnotes

Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026, G.S.R. 120(E) (Feb. 10, 2026) (notified by Ministry of Electronics & Information Technology) [hereinafter IT Amendment Rules, 2026]. ↑
Information Technology Act, 2000, § 87. ↑
Shreya Singhal v. Union of India, (2015) 5 SCC 1. ↑
IT Amendment Rules, 2026, supra note 1, Rules 3 & 4 (inserting SGI-specific due diligence). ↑
IT Amendment Rules, 2026, supra note 1, Rule 2(1)(wa). ↑
Id. ↑
IT Amendment Rules, 2026, supra note 1, Rule 2(1)(wa) proviso. ↑
IT Amendment Rules, 2026, supra note 1, Rule 3(3)(a)(ii) & Rule 4(1A)(c). ↑
Ministry of Electronics & Information Technology, Frequently Asked Questions on the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026 (Feb. 10, 2026). ↑
IT Amendment Rules, 2026, supra note 1, Rule 3(3)(a)(ii) & Rule 3(3)(b). ↑
K.S. Puttaswamy v. Union of India, (2017) 10 SCC 1 (right to privacy under Art. 21). ↑
Modern Dental College & Research Centre v. State of Madhya Pradesh, (2016) 7 SCC 353. ↑
WITNESS, India’s Synthetic Media Rules Build Enforcement on the Wrong Foundation (Feb. 18, 2026). ↑
Information Technology Act, 2000, § 79. ↑
Shreya Singhal v. Union of India, (2015) 5 SCC 1, ¶¶ 118–120. ↑
IT Amendment Rules, 2026, supra note 1, Rule 4(1A). ↑
Ministry of Electronics & Information Technology, supra note 9, at 8–10. ↑
Gautam Bhatia, Guest Post: A Constitutional Critique of the Synthetically Generated Information (IT Rules Amendment), 2026, Ind. Con. L. Phil. Blog (Feb. 11, 2026). ↑

Sneha Awasthi