Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Usually, no. As of 25 October 2026, India has no single rule that makes every creator announce that artificial intelligence (AI) helped make a post, article, image or video. The important new obligation is aimed at intermediaries—services that host, transmit or distribute content. Under amendments to the Information Technology Rules that took effect on 20 February 2026, covered intermediaries must clearly label and attach traceable metadata to certain permissible synthetically generated information (SGI).
That does not make AI content legally safe. Existing rules still apply to impersonation, privacy violations, obscene or sexually explicit material, deception, misinformation and other unlawful conduct. Whether you need a label therefore depends on what you made, how realistic it is, where it is distributed, and whether a sector-specific rule, contract or platform policy adds another requirement.
![AI Law and Policy: [Connected eBook]](https://m.media-amazon.com/images/I/41Y7jOay2eL._SL160_.jpg)
As an Amazon Associate I earn from qualifying purchases.
Is AI disclosure mandatory for every Indian creator?
No. The current national framework does not impose a universal creator-side disclosure for every work that received AI assistance. A person who uses an AI tool for brainstorming, drafting, editing or making an ordinary illustration is not automatically covered by a single “say this was made with AI” rule.
The legal distinction is between a creator’s general use of AI and an intermediary’s duties for covered synthetic information. The amended Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules require intermediaries to make qualifying synthetic material identifiable. That obligation can affect what a creator must provide to a platform, but it is not the same as a statute ordering every individual to place a label on every AI-assisted work.
#1 Best Overall
Voluntary disclosure can still be sensible. A newsroom, school, employer, client contract or platform may demand it, and transparent wording can reduce confusion when a realistic image or voice could otherwise be mistaken for a real event. Treat that as a policy, contractual or trust decision unless a rule specifically applies to your content or sector.
What changed on 20 February 2026?
The Ministry of Electronics and Information Technology (MeitY) says the 2026 amendment was notified on 10 February 2026 and came into force on 20 February 2026. Government explanations describe a system in which intermediaries must ensure clear labelling and traceable metadata for permissible AI-generated content, allowing users to identify synthetic material.
Free tools Windows power users keep installed
One-click scans. No signup required.
MeitY’s FAQ is useful for understanding the change, but it expressly says it is explanatory and “not a legal document” that replaces, amends or alters the Information Technology Act or the Rules. If the FAQ and the gazetted Rules appear to differ, the notified Rules—and any later official amendment—control. The commencement date also matters: advice based only on pre-20 February 2026 guidance may not describe the current intermediary duties.
What counts as “synthetic information” under the Rules?
The Gazette definition is narrower than “anything made with AI.” It covers audio, images and audiovisual information that has been artificially or algorithmically created, generated, modified or altered so that it appears real, authentic or true and portrays an individual or event as indistinguishable from a natural person or a real-world event.
In practical terms, a fabricated video showing a real person doing something they never did, a cloned voice that sounds like a particular individual, or a photorealistic scene presented as a genuine event is much closer to the definition than a clearly fantastical illustration or a lightly edited photograph. The test focuses on the impression created for viewers, not simply on whether an AI button was used.
| Type of work | How it relates to the SGI definition | What to check |
|---|---|---|
| Photorealistic face or event generated from a prompt | Likely within scope if presented as real or indistinguishable from reality | Could a reasonable viewer mistake it for an authentic person or event? |
| Voice clone or synthetic speech imitating an identifiable person | Likely within scope when it appears to be that person’s genuine voice | Does the presentation imply the person actually spoke? |
| AI video combining generated and real footage | Potentially within scope if the alteration changes the apparent reality of a person or event | Identify the materially altered portion and the context supplied to viewers |
| Clearly fictional animation, game asset or imaginative artwork | Usually outside the definition when it is not represented as real | Do captions, editing or surrounding claims nevertheless present it as factual? |
| Noise reduction, colour correction or compression | Excluded when the change does not materially alter meaning | Confirm that the underlying message and depiction remain materially intact |
Which edits and documents are expressly excluded?
The Gazette excludes routine, good-faith work that does not materially alter meaning. The exclusions are important because many ordinary software features now use machine-learning models without creating a deceptive synthetic person or event.
Rank #2
- Formatting, enhancement, technical correction, colour adjustment, noise reduction, transcription and compression, when they do not materially alter meaning.
- Ordinary preparation of documents, presentations, PDFs, educational or training materials, research outputs, and illustrative or template content, provided no false document or false electronic record is created.
- Changes that improve accessibility, clarity, quality, translation, description, searchability or discoverability, when they do not manipulate a material part of the underlying content.
An exclusion is not a licence to mislead. For example, translating a genuine interview is different from generating a quotation that the interviewee never made. Improving a scan is different from altering a signature or date to create a false record. The material-change and false-record limits are what keep routine production work separate from deceptive synthetic information.
Who has to apply the label: the creator or the platform?
The official materials support a platform or intermediary compliance duty. An intermediary that hosts, shares or distributes qualifying SGI must use clear labelling and traceable metadata so users can identify it. The Rules do not establish a blanket instruction that every individual creator must independently invent and attach a legal label before publishing any AI-assisted work.
In practice, a service may ask uploaders to answer an AI-use question, select a disclosure switch, or preserve provenance data supplied by a creation tool. Those are ways for the intermediary to meet its obligation. If you refuse a required platform declaration, the service may restrict distribution under its terms even though the national law is not a universal creator-label statute.
| Publishing situation | Likely legal focus | Practical action |
|---|---|---|
| You post a realistic synthetic image on a social or video service | The hosting intermediary’s SGI labelling and metadata process | Use the service’s AI-content declaration accurately and retain the original project details |
| You send an AI-assisted text draft directly to a client | Contract terms, professional rules and the truthfulness of the deliverable | Check the engagement agreement; disclose if it requires disclosure or if authorship matters to the recipient |
| You publish a clearly fictional illustration | Whether the presentation could mislead, plus any platform policy | Make the fictional context clear when a realistic presentation could confuse viewers |
| You use AI for translation, captions or accessibility | Whether the process materially changes the underlying content | Review for errors and disclose where a client, regulator or platform requires it |
| You create a manipulated clip of a real person | SGI handling plus impersonation, privacy and other applicable offences | Do not publish unlawful or deceptive material; a label does not cure the underlying harm |
What legal risks apply even when no disclosure rule covers the work?
India’s technology laws are intended to apply without giving AI-generated material a special exemption. MeitY has clarified that the Information Technology Act does not distinguish information generated with AI tools from information generated by users when addressing cyber offences and related obligations.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Depending on the facts, the following conduct can lead to removal requests, platform action, loss of intermediary protections, or civil or criminal consequences under the relevant law:
- Impersonation and identity misuse: a fabricated voice, face or message that falsely appears to come from a real person.
- Privacy violations: using someone’s personal information, likeness or intimate imagery without a lawful basis or consent.
- Obscene or sexually explicit material: including synthetic sexual content involving an identifiable person.
- Deception and misinformation: presenting a fabricated event, document, quotation or financial claim as genuine.
- Other unlawful communications: content that breaches the Information Technology Act or Rules, the Digital Personal Data Protection Act, the Bharatiya Nyaya Sanhita, or another applicable statute.
A disclosure label is not a defence to unlawful conduct. Calling a non-consensual intimate deepfake “AI-generated” does not remove privacy, obscenity or other liability, and identifying a fabricated scam message does not make the scam lawful.
How can a creator decide whether to disclose AI use?
Use a two-part check: first classify the output under the SGI definition; then check the distribution context for additional rules. The following sequence keeps legal duties separate from good-practice transparency.
- Describe the output, not the tool. Record whether it is text, audio, an image, video or a combination, and whether an identifiable person or real event is depicted.
- Test the realism claim. Ask whether an ordinary viewer could reasonably treat the material as an authentic person or event. A realistic depiction presented as factual raises more concern than an obviously fictional one.
- Check for a material change. If the process only formats, translates, enhances, compresses, captions or improves accessibility without changing a material part of the underlying content, it may fit an express exclusion.
- Identify the intermediary. If you will upload or distribute through a platform, read its current AI-content and provenance requirements. The platform may collect the information needed for its own labelling and metadata duties.
- Screen for unlawful effects. Review consent, privacy, impersonation, sexual-content, fraud and misinformation risks. Stop the publication if the underlying act is unlawful; disclosure cannot repair it.
- Check sector and contract rules. Advertising, elections, news, education, cinema, government communications and regulated professions may impose additional standards. A client, employer or commissioning editor can also set a stricter contractual policy.
- Keep an audit record. Preserve prompts, source files, permissions, edits and the final disclosure decision. This helps answer a platform, client or regulator’s question about how a realistic work was made.
This process may lead you to disclose voluntarily even where national law does not compel a creator label. Use precise wording such as “AI-generated image” or “AI-assisted translation” rather than implying that every part of the work was machine-made when humans supplied substantial reporting, editing or review.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →What should an intermediary or publisher do?
An intermediary handling qualifying SGI should build the legal requirement into its product and moderation workflow. At minimum, that means identifying covered synthetic audio, images and audiovisual material, displaying a clear notice that ordinary users can understand, and retaining traceable metadata or equivalent provenance information as required by the Rules.
Policies should distinguish realistic synthetic depictions from routine editing and accessibility work. They should also explain what happens when a user disputes a label, when metadata is missing, or when content appears unlawful regardless of its synthetic status. Human review remains important for borderline cases because the definition turns on material alteration and the impression of authenticity, not merely on a file’s technical origin.
Publishers that are not intermediaries should still set an editorial standard. Require disclosure where readers could reasonably be misled, document consent for identifiable likenesses, and state whether a human verified factual claims. Those controls are risk-management choices unless another law, contract or sector code makes them mandatory.
How does the answer differ for common AI-assisted content?
AI-written articles, posts and scripts
The 2026 SGI definition is directed mainly at realistic synthetic audio, images and audiovisual information. Text produced or edited with an AI assistant is not automatically subject to a universal public label under the materials described here. Accuracy, plagiarism, defamation, confidentiality, copyright and any editorial or contractual disclosure rule still matter.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsRank #4
Images and video that look like real events
A photorealistic scene or altered clip can fall within SGI when it is made to appear authentic. Uploading it to an intermediary brings the platform’s labelling and metadata process into play, while depicting a real person without consent can create separate legal exposure.
Voice cloning and synthetic audio
Audio is expressly included in the definition. A cloned voice presented as a genuine statement should be treated as high risk: follow the intermediary’s declaration process and obtain the permissions needed for the person’s likeness or voice.
Translation, captions and accessibility tools
Translation, description, searchability, discoverability and accessibility improvements are among the listed exclusions when they do not manipulate a material part of the underlying content. Review output for mistranslation or invented content, because an error that changes meaning can move the work outside that safe category and may create ordinary liability.
Documents, PDFs and educational materials
Ordinary preparation of these materials is excluded when it does not create a false document or false electronic record. Altering a certificate, invoice, research result, signature or date to make a false record is a different case and can trigger laws that apply regardless of the AI tool used.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →What if a platform, client or regulator asks for a label?
Follow the stricter applicable requirement. A platform’s terms can condition posting or monetisation on an AI declaration; a client contract can require disclosure even when national law does not; and a sector regulator can impose its own format or approval process. Keep a copy of the rule or contract clause so you can distinguish a private condition from a statutory duty.
For disputes, start with the exact content and context: what was generated or altered, whether it was presented as real, who distributed it, and what harm is alleged. The notified Rules, not a generic social-media explanation, should be used to assess an intermediary’s legal obligation. For a high-stakes publication, obtain advice from an Indian lawyer familiar with the relevant sector.
Does the MeitY FAQ itself create a disclosure obligation?
No. MeitY says the FAQ responds to general questions and is not a legal document replacing or changing the Information Technology Act or the Rules. It is an interpretive aid; the gazetted text and any applicable sector-specific instrument determine enforceable duties.
The same caution applies to older advisories. A March 2024 advisory summary focused on unreliable or under-tested AI models, recommending appropriate reliability labelling and user notices such as consent pop-ups. That guidance does not turn every later AI-assisted creation into a universally labelled work, and it should not be read in isolation from the 2026 amendment.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchPC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Frequently Asked Questions
Does using ChatGPT for a first draft require an AI label in India?
Not as a universal national rule merely because an AI assistant helped with drafting. Check the publication’s editorial policy, your client or employer agreement, and any professional or sector-specific requirement; those can be stricter than the baseline law.
Are deepfake labels required by Indian law?
Realistic synthetic audio, images and video are the kinds of material covered by the 2026 SGI framework, so an intermediary distributing them must provide clear labelling and traceable metadata. The exact label may be applied through the service’s upload or moderation system rather than by a creator adding a legally prescribed phrase by hand.
What should I do if an AI tool changes a genuine document?
Do not rely on an AI disclosure to legitimise a false record. Preserve the original, verify every material field and obtain legal advice before publishing or submitting anything that could be treated as an altered certificate, invoice, signature, date or official record.
Which rule wins if a MeitY FAQ and the gazetted Rules seem inconsistent?
The notified Rules and subsequent legally valid amendments control; the FAQ is explanatory and expressly non-substitutive. For a regulated or high-risk publication, confirm the current gazetted text and any sector-specific instrument rather than relying on a summary.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
