AI-generated content rules in India: labels, deepfakes, platforms
AI-generated content rules in India sit in the IT Rules, 2021, as amended from February 20, 2026. Platforms that let users make or share realistic synthetic audio, images or video must stop unlawful kinds, such as deceptive deepfakes, and label the rest, with metadata where technically feasible. Large platforms must also ask users to declare synthetic content, then check it.
Published and updated
What counts as synthetically generated information?
Synthetically generated information is audio, images or video that a computer creates or alters so it looks real and could pass for a real person or event (rule 2(1)(wa) of the IT Rules). The FAQ from MeitY, the Ministry of Electronics and Information Technology, names deepfakes and cloned voices (Q5). Routine edits, accessibility tools and good-faith documents are left out.
MeitY, the Ministry of Electronics and Information Technology, added the definition with G.S.R. 120(E), opens in a new tab, in force from February 20, 2026. It leaves out content arising from (rule 2(1)(wa), proviso, opens in a new tab):
- Routine or good-faith edits, such as color correction, noise reduction, transcription or compression, that don't materially alter, distort or misrepresent the substance, context or meaning (clause (a))
- Routine or good-faith work on documents, slides, PDFs, training material and research, so long as no false document or false electronic record results (clause (b))
- Tools used only for accessibility, clarity, quality, translation, description, searchability or discoverability, without generating, altering or manipulating any material part of the content (clause (c))
MeitY's FAQ of February 10, 2026, opens in a new tab gives examples. Sharpening a photo without changing what it shows, compressing a video and adding subtitles are not synthetic. An AI-made fake certificate, ID or official letter falls outside the exceptions (Q6).
Text alone isn't synthetically generated information, the FAQ says, since the definition covers only audio, visual and audio-visual content (Q8). A chatbot's article spreading false rumors may still be unlawful information under the general rules. The FAQ adds that it isn't a legal document (introduction, note).
G.S.R. 120(E) says nothing about content posted before February 20, 2026, the day it took effect (rule 1(2)). MeitY's FAQ doesn't address it either.
Which services must follow the synthetic content rules?
The synthetic content rules bind intermediaries whose service may enable, permit or facilitate making, changing, publishing or sharing synthetic content (rule 3(3)(a)). MeitY's FAQ lists AI image and video tools, voice cloning tools, and platforms that publish or spread synthetic media, among others (Q20).
An intermediary is anyone who receives, stores or transmits a record for another person, or provides a service for it (IT Act, section 2(1)(w), opens in a new tab). Whether an AI service is one turns on the nature of the service, its functions and the applicable law, the government told the Lok Sabha on July 29, 2026, opens in a new tab.
Rule 3(3)(a) names no exception for business or enterprise tools. Among the documents read to October 31, 2026, the government has issued no list of AI services that are intermediaries.
The duties every intermediary carries are in IT Rules 2021: what India asks of platforms and intermediaries.
How must AI-generated content be labeled?
Lawful synthetic content must carry a prominent, easily noticed label, or, for audio, a prominent spoken disclosure at the start (rule 3(3)(a)(ii)). Where technically feasible, it must also carry permanent metadata or another provenance mechanism, with a unique identifier for the platform's tool that made it. The platform may not let anyone change, hide or remove either (rule 3(3)(b)).
MeitY's FAQ gives two examples: a visible "synthetically generated" notice on a video, and audio that opens with "This audio is synthetically generated" (Q22). Platforms shouldn't offer "remove watermark" or "export without metadata" features (Q23).
The rule in force sets no size or screen share for the label.
A revised draft MeitY published on April 21, 2026, opens in a new tab would keep the label continuously and clearly visible for the whole duration of the content (para 3). Its start date is blank (rule 1(2)). It's a proposal, and among the documents read to October 31, 2026, it hadn't become law.
Which synthetic content must platforms stop?
A platform covered by rule 3(3) must use reasonable and appropriate technical measures, automated tools included, so users can't make or share synthetic content that breaks any law (rule 3(3)(a)(i)). The rule names four kinds: child sexual abuse and intimate or obscene imagery, false documents, explosives and weapons content, and deepfakes that deceive about a real person or event.
The four kinds (rule 3(3)(a)(i), opens in a new tab, items (I) to (IV)):
- Child sexual exploitative and abuse material, non-consensual intimate imagery, and obscene, pornographic, sexually explicit or privacy-invading content, bodily privacy included (item (I))
- False documents or false electronic records (item (II))
- Content on preparing, developing or procuring explosives, arms or ammunition (item (III))
- Content that falsely shows a real person's identity, voice, conduct, action or statement, or falsely shows an event as having happened, in a way likely to deceive (item (IV))
Once a platform learns of a violation, it must act quickly and appropriately (rule 3(1)(cb); MeitY's FAQ, Q13). That can include removing the content and suspending or ending the account, without destroying evidence (rule 3(1)(ca)(ii)). In line with the law, it may have to disclose the user's identity to a victim who complains, or to someone acting for one.
What must large platforms check for synthetic content?
A significant social media intermediary, one with more than 50 lakh (5 million) registered users in India (rule 2(1)(v); S.O. 942(E), opens in a new tab), must screen uploads before they go live. It asks users to declare synthetic content, checks the declaration with technical measures, and labels what's confirmed (rule 4(1A)). A user's declaration alone isn't enough, MeitY's FAQ says (Q26).
1. Ask the user to declare
Before display, upload or publication, require the user to say whether the content is synthetic (rule 4(1A)(a), opens in a new tab). The FAQ's example is an "AI-generated: Yes/No" question (Q24).
2. Check the declaration
Verify it with appropriate technical measures, automated tools included, given the content's nature, format and source (rule 4(1A)(b)).
3. Label what is confirmed
Show confirmed synthetic content with a clear and prominent label or notice (rule 4(1A)(c)).
A large platform that knowingly permits, promotes or fails to act on synthetic content that breaks the Rules is deemed to have failed its due diligence (rule 4(1A), proviso).
Do creators have to label what they post?
Not directly: the platform must make sure synthetic content carries a label (IT Rules, rule 3(3)(a)(ii)). The duty reaches creators through it: a large platform asks you to declare synthetic content before it goes up, then checks and labels it (rule 4(1A)). Platforms that let you make synthetic content must also warn you of penalties for misuse (rule 3(1)(ca)).
MeitY's FAQ says labeled satire and creative synthetic works that break no law may be permitted, so long as they're declared or labeled (Q7). An agency posting for a client makes the same declaration as any other user (rule 4(1A)(a)).
What happens when the rules are broken?
A platform that fails to observe the IT Rules loses its section 79 protection from liability for users' content. It can also be punished under the Information Technology (IT) Act, the Bharatiya Nyaya Sanhita and other laws (rule 7). Users face the penalties those laws set. The Rules make platforms warn users of them, but set none of their own.
Section 79 of the IT Act, opens in a new tab is that protection, often called safe harbor. It's also lost if the platform took part in the unlawful act, or doesn't remove material quickly after actual knowledge or a government notice (section 79(3)).
Removing content in line with the Rules, even with automated tools, doesn't breach the conditions of section 79(2)(a) or (b) for that protection (rule 2(1B)).
How IndusDesk helps with AI-generated content rules
We watch the authorities, tell you what each change means for your products and plans, represent you while a decision is still open, and carry the work through, with every step on record in your workspace.
Frequently asked questions
How fast must a platform remove content now: 3 hours, 2 hours or 36?
Each applies to a different trigger, all since February 20, 2026 (G.S.R. 120(E), opens in a new tab), though MeitY's 2021 FAQ, opens in a new tab still gives the old limits. It's 3 hours after a court order or an authorized officer's reasoned intimation (rule 3(1)(d)), and 2 hours after a complaint about intimate images or impersonation, morphed photos included (rule 3(2)(b)). A complaint asking to remove content rule 3(1)(b) lists gets 36 hours, unless it's an ownership, IP or general-law claim (rule 3(2)(a)(i), proviso); the IT Rules guide has the rest.
Are the synthetic content rules India's AI law?
No: India applies its existing laws to AI, the IT Rules among them. The government calls the IT Act technology-neutral, applying whatever the technology, AI included (Lok Sabha, July 29, 2026, opens in a new tab). The India AI Governance Guidelines, opens in a new tab, released on November 5, 2025, say a separate law to regulate AI isn't needed at this stage (overview of the framework), and AI governance in India has the detail.
Do you need government permission to launch an AI model or chatbot in India?
The India AI Governance Guidelines, opens in a new tab say proposed AI governance frameworks should avoid compliance-heavy requirements, such as mandatory approvals or licensing conditions, unless deemed necessary (Part 4). Existing law still applies: an intermediary whose tool makes realistic audio, images or video must stop unlawful synthetic content and label the rest (rule 3(3)). Text-only output isn't synthetic content (MeitY's FAQ, Q8), and the AI governance guide goes further.
Sources
- Ministry of Electronics and Information Technology: Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021, as updated on February 10, 2026, rules 2(1)(v), 2(1)(wa), 2(1B), 3(1)(ca), 3(1)(cb), 3(1)(d), 3(2), 3(3), 4(1A) and 7, opens in a new tab
- Ministry of Electronics and Information Technology, G.S.R. 120(E), February 10, 2026: Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026, rules 1(2), 2 and 3, opens in a new tab
- Ministry of Electronics and Information Technology, February 10, 2026: Frequently Asked Questions on the IT (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026, introduction (note) and Questions 5 to 8, 13, 20, 22 to 24 and 26, opens in a new tab
- Ministry of Electronics and Information Technology, draft (F. No. 2(5)/2025-CyberLaws), April 21, 2026: Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Second Amendment Rules, 2026, rule 1(2) and para 3, opens in a new tab (PDF, 143 KB)
- Ministry of Electronics and Information Technology, S.O. 942(E), February 25, 2021: threshold for a significant social media intermediary, opens in a new tab
- The Information Technology (Amendment) Act, 2008 (Act No. 10 of 2009): sections 2(1)(w) and 79, opens in a new tab
- Lok Sabha Unstarred Question No. 1647, answered July 29, 2026: whether generative AI systems are intermediaries eligible for safe harbor protection, opens in a new tab
- Lok Sabha Unstarred Question No. 1645, answered July 29, 2026: legislation for artificial intelligence, opens in a new tab
- Ministry of Electronics and Information Technology, released November 5, 2025: India AI Governance Guidelines, overview of the framework and Part 4 (guidelines for regulators), opens in a new tab
- Ministry of Electronics and Information Technology, October 2021: Frequently Asked Questions on the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021, Question 4(e), opens in a new tab