This Article Was Written By Sakshi Poojary. This article, written by the Indian government, has introduced new rules to deal with the threat of fake media. These rules, which are part of the Information Technology Act, went into effect on February 20, 2026.

Introduction

A fake video of a famous movie star spread like wildfire on social media in India back in November 2023. Millions watched and shared it quickly, but by the time someone in charge stepped in to set the record straight, the damage was already done. The truth took a lot longer to catch on than the lie, which is pretty scary. The incident is frequently cited as the turning point in Indian public law on deepfakes. Until now, people who were victims of certain crimes online had limited options to seek help. They could use sections 66C and 66D of the Information Technology Act of 2000, which deal with identity theft and impersonation. There was also section 66E, which is about privacy violations, and sections 67 and 67A, which cover obscene content.

Additionally, they could use general laws against defamation and forgery, which are now part of the Bharatiya Nyaya Sanhita,2023. However, these laws were not specifically designed to handle issues related to synthetic media, which has become a growing concern.

So the Indian government has come up with new rules to deal with the threat of fake media. These rules, which are part of the Information Technology Act, went into effect on February 20, 2026.

The new rules include a term called “synthetically generated information,” which refers to audio, video, or a mix of both that’s created artificially or changed by algorithms to look real. This doesn’t include normal editing or fixing minor technical issues.

If something is synthetic, it needs to be clearly labeled. To address the issue of non-consensual intimate images on social media, important intermediaries must take a few key steps. First, they need to get a declaration from the person who uploaded the content and then use technical tests to verify the information provided. When it comes to non-consensual intimate images, the time allowed to act on a court order. Instead of fifteen days, grievances must be settled in seven. A platform runs the risk of losing the safe harbor provided by section 79 of the IT Act if it knowingly allows or ignores content that violates the law.

Few would contest the goal. A convincing fake can do genuine and often irrevocable harm, whether it be sexual, financial, or political. A person’s face and voice are among their most private possessions. The method is the challenge. Three hours is not enough time to distinguish a malevolent fabrication from a humorous edit, parody, or an unfavorable but authentic recording. It seems sensible for a platform that could lose its immunity to prioritize removal above investigation, and observers have also noted that the Rules do not specify how automated detection reliability should be assessed. This essay poses the question of whether the framework can safeguard digital identity without undermining the Constitution’s Article 19(1)(a). It contends that the proportionality standards established by the courts are more important in determining the outcome than the wording of the Rules.

It is instructive to compare with other jurisdictions. China demands visual watermarking of synthetic media, California has focused mostly on election-related fakes, and the European Union has selected a risk-based model under its AI Act. The model used in India is distinct. It is designed around-the-clock, combining traceability obligations with extremely brief takedown timeframes, and it places the whole responsibility of compliance on intermediaries rather than content creators. This ruling speeds up enforcement, but it also puts the platform – rather than the court- first and frequently the last arbiter of what the public can view.

Case Laws

The rulings below demonstrate how Indian courts have addressed the long-standing constitutional issues brought up by the 2026 Rules.

Any analysis of intermediary liability has to start with the Supreme Court’s judgment in Shreya Singhal v. Union of India, (2015) 5 SCC 1. The Supreme Court has restricted Section 79(3)(b) to “actual knowledge” based on a Court Order or a Government Notice. Article 19(2) is safeguarded by the Court in the sense that S.69A is upheld, given it is accompanied by reasons and other safeguards, whereas the requirement for notice to trigger a three-hour clock to delete information was struck down in strikes 66A from the books. The three-hour period for a court order or notice as specified in the 2026 Rules applies accordingly. The strain arises when a platform has to respond to a user complaint or its own detection because the need to delete content swiftly conflicts with a standard based on factual knowledge.

Justice K.S. Puttaswamy (Retd.) v. Union of India is a landmark case that recognized the right to privacy, dignity, and informational autonomy under Article 21 of the Constitution. This case made it clear that individuals have a right to be protected from deepfakes and other violations of their privacy.

When it comes to limiting speech to stop harassment, the rules must be fair. They must follow the law. There must be a valid reason for the government to step in. The Supreme Court recently gave a standard for this in the Modern Dental College v. State of Madhya Pradesh case. That standard was then used in another case, Anuradha Bhasin v. Union of India. The Supreme Court applied the principle of proportionality in that case to decide how online speech should be regulated. The court said that any restriction, on speech must be reasonable and must not go beyond what is necessary.

The Supreme Court set a standard for proportionality and then used it to decide a case about online speech, called Anuradha Bhasin v. Union of India. In this case, the Court applied the same rules to online communication. So, even if you limited speech for just three hours to help someone in trouble, it still wouldn’t be enough if there wasn’t a person making decisions and using their judgment.

Kunal Kamra v. Union of India (Bombay High Court, 2024) is a useful warning from the same rulebook. The High Court struck down the 2023 amendment to Rule 3(1)(b)(v), which allowed a government Fact Check Unit to identify “fake, false or misleading” information about the Union, as inconsistent with Articles 14 and 19. The lesson is that the State cannot be the sole judge of truth. The 2026 amendment avoids that particular defect, since it is tied to unlawful content and not to official versions of fact. Even so, the same vice can return quietly if “unlawful” is read broadly by those who issue the notices.

The High Courts have been prompt and creative when it comes to identity protection.In the case of Anil Kapoor versus Simply Life India in the Delhi High Court in 2023 the court decided to stop content from being shared and stopped anyone from using the actors name, picture, and voice, even with the help of artificial intelligence. In another case, Arijit Singh versus Codible Ventures LLP, in 2024 the Bombay High Court stopped AI tools that copied the singer’s voice. The Delhi High Court also protected Aishwarya Rai Bachchan from deepfake abuse in 2025.  These are temporary orders that have primarily benefited public people who have the resources to file lawsuits. The two-hour clock for personal imagery may prove to be the most justifiable aspect of the new framework because the average woman whose face appears on an obscene film has no such recourse.

Conclusion

The 2026 Rules are commendable for identifying the issue, describing it,

and realizing that a deepfake that circulates for a day has already done the majority of its damage.

Since labeling and provenance requirements inform the listener

rather than suppressing the speaker, they are, in theory, the least intrusive tools available. It is also simple to argue for the quick removal of personal imagery.

The consistent compression of time is the weak point. Satire, political commentary,

And journalism will pay the price for automated over-removal,

which is encouraged by a three-hour deadline linked to an imprecise concept of unlawfulness,

with impunity as the price of error. A better calibrated design is required for proportionality.

Platforms should offer a brief but genuine window for human assessment of content

that is considered borderline, notices should clearly state their legal basis,

and erroneous removal should have a remedy that is just as easily accessible as the takedown itself. It would also be beneficial for the government to release explicit guidelines on parody and the evaluation of detection accuracy. A requirement that only the biggest corporations can comply with will

subtly entrench smaller platforms, which are unable to staff a 24-hour review desk.

India’s decision to regulate early is justifiable.

In the courtroom, where the principles of Shreya Singhal, Puttaswamy,

And Anuradha Bhasin will determine whether preserving digital identity and

Defending free speech is considered as rivals or as components of one constitutional guarantee,

the question of whether it proves sensible will be resolved less in the Gazette.

FAQs

Q1. What is a deepfake under the 2026 amendment?

The Rules use the term “synthetically generated information”.

This refers to audio, video, or a mix of both that’s been made or

changed using artificial means, like algorithms, to make it look real.

Simple tweaks, like editing or translating, don’t count – we’re talking about more

complex changes that can make something fake seem genuine.

Also, fixing technical issues in a way that’s trying to help is okay,

but this rule doesn’t apply to content that’s just text.

Q2.  How quickly must platforms remove unlawful content now?

A new rule is in place: if a court orders it or the government says so,

things need to happen faster – within three hours, not thirty-six.

This is especially true for really sensitive stuff like private photos or

videos that were shared without permission – those need to be taken care of within two hours, not twenty-four. And if a user has a problem, it must be solved within 7 days.

Q3. Are all AI-generated videos prohibited?

When it comes to synthetic content, there are rules to follow. If the content is lawful, it can stay online, but it needs to have a clear label and information about where it came from. On the other hand, content that’s against the law, like intimate images shared without consent, must be removed

Q4. Can a platform lose its legal immunity under the Rules?

If someone helps spread fake content on purpose, or does nothing to stop it, they can get in trouble. They might not be protected by the law anymore, specifically Section 79 of the IT Act, because they didn’t do their job properly.