Quick answer
India regulates deepfakes through amendments to the IT (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021. MeitY's 22 October 2025 draft defined "synthetically generated information" and required visible labels and metadata; final amendments were notified in February 2026 and changed several draft specifics. Intermediaries that fail due diligence risk losing safe harbour under section 79 of the IT Act.
- Deepfake rules in India sit inside the IT Rules, 2021, not in a standalone AI law.
- The October 2025 draft defined synthetically generated information and proposed labels covering at least 10% of visual area or the first 10% of audio.
- Final amendments were notified in February 2026 and changed several draft specifics; read the notified Gazette text before building to any percentage or deadline.
- Non-compliant intermediaries risk losing section 79 safe harbour, exposing them to liability for user content.
- Deepfakes of real people are also personal data processing, so DPDP duties apply from 13 May 2027.
Who this is for: Social media platforms and significant social media intermediaries, GenAI image, video and voice tool builders, Content and trust-and-safety teams, Media and marketing agencies, Legal and policy teams at intermediaries
What are India's deepfake and synthetic content rules?
India has no standalone deepfake statute. Instead, the Ministry of Electronics and Information Technology (MeitY) has used the IT (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 — the rules that set due diligence duties for intermediaries under the IT Act — to address AI-generated content.
The October 2025 draft
On 22 October 2025, MeitY published draft amendments that:
- defined "synthetically generated information" — content artificially or algorithmically created, generated, modified or altered so that it appears authentic;
- required such content to carry visible labels and embedded metadata;
- proposed that labels cover at least 10% of the visual area, or for audio the first 10% of the duration;
- required significant social media intermediaries to obtain a user declaration on whether uploaded content is synthetic, and to verify that declaration using reasonable technical measures.
The February 2026 final amendments
Final amendments were notified in February 2026. The final text changed several draft specifics. This page deliberately does not restate final percentages, label formats or compliance windows, because builders should work from the notified Gazette text itself. Download it from egazette.gov.in and check MeitY's site for any FAQs or advisories issued with it.
What has not changed is the structure: synthetic content is now a named category inside intermediary due diligence, and labelling plus provenance is the expected control.
Who has to comply
The IT Rules impose duties on intermediaries — entities that receive, store or transmit content on behalf of others. In the synthetic content context that covers three groups.
| Actor | Typical example | Likely duty focus |
|---|---|---|
| Platforms hosting user content | Social networks, video sharing, messaging with public features | Detect, label, act on complaints and orders |
| Significant social media intermediaries | Large platforms above the notified user threshold | Additional duties such as user declarations and verification (per draft; confirm in final text) |
| Tools that generate synthetic content | Image, video and voice generators offered to the public | Labelling and metadata at generation time |
Whether a generative AI tool is itself an "intermediary" depends on how it is offered and the final definitions. A tool that lets users create and share content publicly is more exposed than an internal enterprise model.
Users and advertisers
The rules primarily bind intermediaries, but platforms pass duties to users through terms of service: declaring synthetic uploads, not removing labels, and not impersonating others. Marketers using synthetic spokespeople should expect platforms to require disclosure.
If you sell to Europe as well, the EU AI Act's Article 50 imposes separate marking duties — compare them in AI labelling in India vs EU AI Act Article 50.
Duties that already applied before the amendments
Synthetic content rules were added to an existing due diligence framework. Those base duties still apply:
- Rule 3(1)(b) due diligence. Intermediaries must make reasonable efforts to ensure users do not host or share prohibited content, which includes content that impersonates another person or is patently false and misleading.
- Takedown on orders. Content must be removed or disabled when directed by a court order or an authorised government notification, within the timelines in the IT Rules.
- Grievance redressal. A grievance officer and complaint handling process are required.
- Safe harbour under section 79. An intermediary is protected from liability for third-party content only if it observes due diligence. Failing to comply with the IT Rules can strip that protection.
The practical consequence: losing safe harbour does not create a fixed fine, but it means the platform can be held liable for the deepfake itself — civil claims, criminal complaints and regulatory action become possible.
Why this matters for AI builders
Even before labelling duties, a platform that ignored complaints about non-consensual deepfakes was exposed. The amendments make the expected control explicit and auditable.
How to implement labelling and provenance
Build to the notified text, but these engineering patterns hold under any final specification.
At generation time
- Visible label. Render a persistent overlay on images and video, and a spoken or audible disclosure on audio. Make size and position configurable so you can match the final rule without a redeploy.
- Embedded metadata. Write provenance metadata into the file (for example, content credentials in the C2PA style) identifying it as AI-generated, the tool and a timestamp.
- Watermarking. Add an invisible watermark that survives common transformations, as a backup when metadata is stripped.
At upload time (platforms)
- Ask uploaders to declare whether content is synthetic.
- Run detection: check metadata, watermark signals and classifier scores.
- Where declared or detected, apply the platform label and keep a record.
- Do not let users remove platform-applied labels.
Logging
Keep an audit trail of declarations, detection results, label decisions and takedowns. These logs contain personal data (uploader identity, content of identifiable people), so they fall under DPDP one-year log retention from 13 May 2027 and need Rule 6 safeguards.
Deepfakes and the DPDP Act
A deepfake of an identifiable person is built from, and is itself, personal data. That pulls the DPDP Act into every synthetic media pipeline.
- Ground for processing. Generating a likeness of a real person needs consent under section 6 or a section 7 legitimate use. Training a voice clone on a customer's call recordings, for instance, needs a consent that names that purpose.
- Publicly available exemption is narrow. Section 3(c) excludes data made publicly available by the Data Principal herself or under legal obligation. A celebrity's public photos do not give a tool licence to generate new content of them for any purpose; read DPDP and publicly available data.
- Children. Section 9(3) prohibits tracking, behavioural monitoring and targeted advertising directed at children, and section 9(2) prohibits detrimental processing. Synthetic content of minors is high risk on both counts.
- Security and breach. Face and voice datasets used for generation need Rule 6 safeguards. A leaked training set is a personal data breach reportable under Rule 7.
- Erasure. When consent is withdrawn, section 8(7) requires erasure, including from datasets and, where feasible, fine-tuned artefacts. See erasure in LLMs.
These obligations apply from 13 May 2027 and sit alongside, not inside, the IT Rules.
Common mistakes and what to do now
- Building to the draft numbers. The 10% label size and first-10% audio rule were draft proposals. Hard-coding them risks non-compliance with the final text.
- Metadata only. Metadata is stripped by screenshots and re-encoding. Pair it with a visible label and, ideally, a watermark.
- Treating it as a moderation problem only. Generation tools need controls at creation time, not just platforms at upload time.
- Ignoring voice. Audio deepfakes are covered. Voice cloning products need audible disclosure and consent records.
- Forgetting DPDP. IT Rules compliance does not cover consent, security or erasure duties for the people depicted.
Action list
- Download and read the notified February 2026 amendment from the Gazette.
- Map each obligation to a product surface: generator, upload flow, feed, ads.
- Ship configurable labelling and metadata.
- Add synthetic content categories to your grievance and takedown workflows.
- Record DPDP consent for any real-person likeness and plan for AI product DPDP compliance by 13 May 2027.
Further reading on suhasbhairav.com
- BlogDeepfake risk management for user-generated content products
- BlogContent Moderation vs Policy Enforcement in Enterprise AI
- BlogData Leakage Prevention vs Content Moderation in AI Pipelines
- AI SkillsContent Moderation AGENTS.md Template
- AI AutomationAI Automation for Social Media Management in Marketing
Before 13 May 2027
Compliance checklist
Read the notified amendment
Obtain the February 2026 Gazette notification and list every synthetic content obligation with its deadline.
IT Rules, 2021 (as amended)
Classify your role
Decide whether you are an intermediary, a significant social media intermediary or a tool provider for each product.
IT Rules, 2021
Ship visible labels
Add configurable visible labels for images, video and audio generated or flagged as synthetic.
IT Rules synthetic content amendments
Embed provenance metadata
Write machine-readable provenance into generated files and preserve it through your pipeline.
IT Rules synthetic content amendments
Collect upload declarations
Ask uploaders whether content is synthetic and verify with technical checks where required.
IT Rules (SSMI duties)
Update takedown workflow
Add synthetic content and impersonation categories with clear response timelines.
Rule 3(1)(b)
Record DPDP consent for likeness
Capture consent naming synthetic generation before using a real person's face or voice.
DPDP s.6
Secure training datasets
Encrypt, restrict and log access to face and voice datasets.
DPDP Rule 6
Key dates
MeitY publishes draft IT Rules amendments on synthetically generated information
DPDP core obligations apply to personal data in synthetic media pipelines
Answers
Frequently asked questions
Are deepfakes illegal in India?
Are deepfakes illegal in India?
There is no blanket ban on synthetic content. Deepfakes that impersonate, defame, deceive or harm people can breach existing laws, and intermediaries must follow the IT Rules, 2021, which were amended in February 2026 to cover synthetically generated information, including labelling. Lawful synthetic content must generally be labelled.
What is synthetically generated information under the IT Rules?
What is synthetically generated information under the IT Rules?
MeitY's October 2025 draft defined it as information artificially or algorithmically created, generated, modified or altered using a computer resource in a way that appears reasonably authentic or true. The final February 2026 amendment uses the concept; check the notified text for the exact definition.
Do AI labels have to cover 10% of the image in India?
Do AI labels have to cover 10% of the image in India?
The 10% visual area and first-10% audio duration figures come from the October 2025 draft. The final amendments notified in February 2026 changed several draft specifics. Build to the notified Gazette text, not the draft figures, and keep label size configurable.
What happens if a platform does not label deepfakes?
What happens if a platform does not label deepfakes?
An intermediary that fails to observe IT Rules due diligence risks losing safe harbour under section 79 of the IT Act. Without safe harbour, the platform can be held liable for the user content itself, and it may face directions and other action under the IT Act.
Does the DPDP Act apply to deepfakes?
Does the DPDP Act apply to deepfakes?
Yes, where the content depicts an identifiable person. Creating synthetic media of a real person processes their personal data, so you need consent or a legitimate use, security safeguards, breach reporting and erasure on withdrawal. These DPDP obligations apply from 13 May 2027.
Do enterprise GenAI tools need deepfake labels?
Do enterprise GenAI tools need deepfake labels?
It depends on whether the tool is offered publicly as an intermediary and whether its output is shared publicly. Internal tools are less exposed under the IT Rules, but labelling generated media is still good practice, and EU Article 50 may apply if outputs reach EU users.
Legal references and sources
- IT (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 — Rule 3(1)(b) due diligence; synthetic content amendments (Feb 2026) (official text)
- Digital Personal Data Protection Rules, 2025 — Rule 6 and Rule 7 (official text)
This guide explains the law for engineering and business teams. It is not legal advice; confirm how it applies to your organisation with qualified counsel, and check the Gazette for later notifications or amendments.
About the author

Suhas Bhairav
Systems architect and applied AI practitioner who designs production AI systems for Indian and global companies, with a focus on privacy-by-design under the DPDP Act and the EU AI Act.
Related guides
- India AI & sector rulesAI Labelling in India vs EU AI Act Article 50Side-by-side comparison of India's IT Rules synthetic content labelling and EU AI Act Article 50 transparency duties, with a single build plan that can satisfy both.
- India AI & sector rulesIndia AI Regulation in 2026: The Complete LandscapeIndia regulates AI through existing laws, not a standalone AI Act: MeitY's AI Governance Guidelines, DPDP, IT Rules on synthetic content, CERT-In directions, RBI FREE-AI and SEBI rules.
- Sector guidesDPDP for Social Media IntermediariesSocial media intermediaries with at least 2 crore registered users face three-year inactivity erasure, plus children's, rights, breach and government-request duties under DPDP.
- AI & personal dataFacial Recognition and Biometrics Under the DPDP ActFace recognition, fingerprint attendance and other biometrics are personal data under the DPDP Act. How to choose a basis, protect templates and avoid over-collection.
- AI & personal dataVoice AI and Call Recordings Under the DPDP ActCall recordings, transcripts, voice bots and voiceprints all process personal data. How to handle notice, consent, redaction, retention and vendors under DPDP.
- India AI & sector rulesIndia AI Governance Guidelines (MeitY) Explained for BuildersMeitY's India AI Governance Guidelines (November 2025) explained for builders: the seven sutras, six pillars, proposed institutions, and how to turn them into engineering practice.