Quick answer: Since February 2026, India’s IT Rules require online platforms to clearly label realistic AI-generated or AI-altered audio, images and video (the rules call it synthetically generated information), to attach provenance metadata where technically feasible, and to act much faster on unlawful content. If you are a creator, freelancer or small business, the practical impact is simple: when a platform asks whether your upload is AI-generated or altered, answer honestly, never try to strip labels, and never create realistic AI content that impersonates a real person or event.
Deepfakes, cloned voices and fake ‘news clips’ have become a real problem in India, from investment scams using celebrity faces to family-emergency voice calls. The government responded by amending the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021. This guide explains, in plain English, what changed, who has to do what, and how you can keep publishing AI-assisted content without getting your posts or accounts into trouble.
What changed and when
The Ministry of Electronics and Information Technology (MeitY) notified the amendment as G.S.R. 120(E), dated 10 February 2026. According to MediaNama’s report on the notification, the rules came into force on 20 February 2026, giving platforms a ten-day window to comply. A draft had been released for public comment in October 2025.
The amendment does three big things:
- Defines synthetic content so that everyone knows what needs a label.
- Puts labelling and provenance duties on platforms, especially those that offer AI creation tools and large social media platforms.
- Shortens takedown and grievance timelines across the board.
What counts as ‘synthetically generated information’?
Under the new definition in rule 2(1)(wa), synthetically generated information is audio, visual or audio-visual content that is artificially or algorithmically created, generated, modified or altered using a computer resource, in a way that makes it appear real, authentic or true, and that depicts a person or event so that it is, or is likely to be perceived as, indistinguishable from a real person or real-world event.
The key test is realism. A cartoon-style illustration that nobody would mistake for a photograph is very different from a lifelike video of a politician, a cloned voice of your CEO, or a fake photo of a flood that never happened.
What is excluded
The rules carve out ordinary, good-faith work. Based on the consolidated rules published by MeitY, these are not treated as synthetic content:
- Routine editing that does not materially change meaning, such as formatting, colour correction, noise reduction, transcription or compression.
- Preparing documents, presentations, educational or research material, including illustrative drafts and templates, as long as it does not create a false document or record.
- Accessibility and quality improvements such as translation, descriptions or making content searchable, without materially altering the underlying content.
So brightening a product photo or auto-generating captions is fine. Making a realistic video of a customer who does not exist giving a testimonial is not ‘routine editing’.
What platforms must now do
Platforms that offer AI creation tools
If a service lets users create or alter content with AI, rule 3(3) requires it to use reasonable technical measures to stop users from producing unlawful synthetic content. The rules specifically list:
- child sexual abuse material and non-consensual intimate imagery;
- false documents or false electronic records;
- content about making explosives, arms or ammunition;
- content that falsely depicts a real person or real-world event in a deceptive way.
Permitted synthetic content must be prominently labelled so the label is clearly visible in the display (audio needs a prominently prefixed audio disclosure), and it must carry permanent metadata or other provenance markers, including a unique identifier for the tool used, to the extent technically feasible. Platforms must not let users remove or tamper with these labels or markers.
One notable change from the October 2025 draft: the draft wanted labels to cover at least 10% of the screen or audio duration. MediaNama reports that this fixed size requirement was dropped in the final rules, which now simply require a prominent, easily noticeable label.
Large social media platforms
Significant social media intermediaries (the large platforms) have extra duties under rule 4(1A). Before content is published they must:
- ask users to declare whether the content is synthetically generated;
- use technical measures, including automated tools, to verify that declaration; and
- ensure confirmed synthetic content is clearly and prominently labelled.
This is why you may increasingly see an ‘AI-generated’ or ‘altered content’ toggle when uploading to large platforms. In India, that question is now backed by law.
Faster takedowns and grievance handling
| Obligation | Earlier timeline | New timeline |
|---|---|---|
| Remove unlawful content after actual knowledge (rule 3(1)(d)) | 36 hours | 3 hours |
| Act on complaints about nudity, sexual content or impersonation (rule 3(2)(b)) | 24 hours | 2 hours |
| Resolve general grievances (rule 3(2)(a)(i)) | 15 days | 7 days |
| Resolve content-removal complaints (proviso to rule 3(2)(a)(i)) | 72 hours | 36 hours |
Platforms must also remind users of their rules and the consequences of breaking them at least once every three months, instead of once a year. Services that offer AI tools must specifically warn users that misuse can lead to content removal, account suspension or termination, disclosure of identity to a victim where the law permits, and action under applicable laws.
What this means for creators, freelancers and small businesses
The legal duties sit mainly with platforms. But platforms pass them on to you through upload declarations, automated detection and faster removals. A wrong declaration can mean a removed post, a strike or a suspended account, which can hurt a business page or a client’s ad campaign.
A simple compliance checklist
- Declare honestly. If an upload contains realistic AI-generated faces, voices or scenes, tick the AI or altered-content option the platform offers.
- Do not remove watermarks or metadata. Many AI tools add visible marks or invisible provenance data. Cropping out a label or stripping metadata is exactly what the rules tell platforms to prevent.
- Never impersonate real people. No AI ‘endorsements’ from celebrities, no cloned voice of a competitor, no fake customer reviews in video form.
- Get written consent before creating an AI avatar or voice of a real person, even a team member.
- Add your own disclosure. A short line such as ‘This video uses an AI-generated voiceover’ in the caption builds trust, even when the platform also labels it.
- Keep project files. Save prompts, source footage and tool names. If a post is flagged, you can show how it was made.
- Brief your clients. If you are a freelancer making ads or reels, put AI use and labelling in your proposal and contract.
Examples
- A bakery in Pune uses an AI tool to create a realistic short video of a chef who does not exist. That looks like a real person, so declare it as AI-generated when uploading.
- A coaching institute uses AI to translate a teacher’s real lecture into Hindi subtitles. That is an accessibility use and is generally excluded, but avoid AI lip-sync that makes the teacher appear to say things they did not.
- A freelancer creates an illustrated infographic with AI. It is clearly not mistaken for reality, so it is unlikely to fall under the definition, though disclosing AI use is still good practice.
How to protect yourself as a viewer
Labels help, but scammers do not follow rules. Treat any urgent video or voice request for money with suspicion, verify through a second channel, and report fake content on the platform. Our guide on stopping AI voice scams explains a family code-word system that works well. If you are new to AI tools, read what an AI agent is to understand where automated content comes from, and browse more guides in our AI Awareness section.
The simplest rule: if a reasonable viewer could mistake your AI content for a real person or a real event, label it.
Criticism and open questions
Not everyone is happy with the rules. Digital rights groups such as the Internet Freedom Foundation have argued that very short takedown windows may push platforms to remove lawful content to be safe, and that automated verification of user declarations can be error-prone. These debates may lead to further clarifications, so check MeitY updates if you publish AI content at scale.
FAQ
Do I need to label every image I edit with AI?
No. Routine edits that do not materially change meaning, such as colour correction or noise removal, are excluded. The rules target realistic content that could be mistaken for a real person or event.
Is it illegal to make deepfakes in India now?
The amendment does not ban AI content. It requires labelling and requires platforms to block unlawful categories such as impersonation, false documents and non-consensual imagery. Creating deceptive or harmful deepfakes can also attract action under other laws.
Does the label have to cover 10% of the screen?
No. That was in the October 2025 draft. The final rules require a prominent, easily noticeable label without a fixed percentage, according to MediaNama’s analysis.
What happens if I declare wrongly?
Platforms can remove the content, restrict or suspend your account, and in serious cases disclose your identity to a victim as permitted by law. Honest declaration is the safest path.
Sources
- MeitY – IT (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021, consolidated text updated 10.02.2026: definition of synthetically generated information, exclusions, rule 3(3) and 4(1A) duties, takedown timelines and three-month user reminders.
- MediaNama – MeitY Notifies New Amendments to IT Rules on Synthetic Media: notification and effective dates, removal of the 10% label-size requirement, summary of timeline changes.
- Internet Freedom Foundation – IT Intermediary Amendment Rules, 2026 contradict their purpose: criticism of the amendment.
Disclaimer: This article is general information, not legal advice. Rules and platform policies can change; consult a qualified lawyer for advice on your specific situation.