New IT Rules on Synthetic Media in India: Takedown Deadlines Content Syndicators Must Follow
SuprContentIn shortIndia’s 2026 amendments to the Information Technology Rules, 2021 introduce faster content-removal deadlines and new duties for synthetically generated information.
On this page14 sections
For Indian content syndicators, publishers, aggregators, distribution platforms, and intermediary services, the main operational issue is knowing which legal clock applies to which request. The current rules require action within 3 hours for qualifying court orders or reasoned government intimations, 2 hours for specified victim complaints involving intimate content or electronic impersonation, 36 hours for certain other removal grievances, and 7 days for general grievance resolution. The synthetic media amendments were notified on February 10, 2026, and took effect on February 20, 2026.
The distinction between these deadlines matters because the two-hour rule is not a universal deadline for every deepfake, and the three-hour rule does not begin whenever any user reports AI-generated content. Content syndicators need a classification process that identifies the legal role of the service, the type of synthetic content involved, the source of the request, and the exact rule that starts the response clock.
The Three-Hour Takedown Rule Applies After Specific Forms of Actual Knowledge
The three-hour deadline applies when an intermediary receives actual knowledge through a qualifying court order or a reasoned written intimation from the appropriate government or an authorized agency. The 2026 amendment reduced this period from 36 hours to three hours. A normal user complaint, automated deepfake alert, editorial concern, or general email does not by itself start this particular three-hour clock.
Rule 3(1)(d) is broader than synthetic media. It covers unlawful information that an intermediary must remove or restrict access to after receiving the specified form of actual knowledge. Synthetic media can enter this process when an AI-generated or manipulated item falls within content prohibited under applicable law.
This makes intake classification one of the most important operational tasks for a syndication business. A team receiving a notice should immediately determine:
- Whether the sender is a court, authorized government officer, ordinary user, affected individual, business partner, or internal reviewer.
- Whether the communication identifies a specific URL, account, post, file, asset, or other content identifier.
- Whether a government communication is the reasoned written intimation contemplated by the rules.
- Whether the material is hosted, stored, transmitted, republished, embedded, cached, or merely linked through the service.
- Whether another shorter or longer grievance deadline applies.
The three-hour period creates little room for a notice to remain unclassified in a general support queue. A syndication service that operates across weekends, nights, regional desks, or multiple publishing systems therefore needs an escalation process capable of identifying a qualifying legal notice as soon as it arrives.
The official rule text is especially significant for freshness. Some later online summaries continued describing the three hours as a proposed change. The notified Gazette dated February 10, 2026 already substituted three hours for 36 hours, with the amendment taking effect on February 20, 2026. Operational policies should therefore be built from the notified rule text rather than older draft descriptions.
The Two-Hour Rule Is Narrower Than a General Deepfake Takedown Requirement
The two-hour deadline applies to complaints involving specified personal harms, including material exposing a person’s private area, nudity, sexual acts or conduct, and electronic impersonation that can include artificially morphed images. The amendment reduced the previous 24-hour period to two hours. It should not be read as a two-hour removal requirement for every form of synthetic media.
This distinction changes how deepfake complaints should be routed.
A manipulated political speech, AI-generated disaster image, synthetic product demonstration, cloned celebrity voice, fabricated business announcement, and non-consensual intimate deepfake may all involve synthetic media. Still, they do not automatically enter the same complaint process.
A two-hour workflow should be reserved for material that falls within the categories specified by Rule 3(2)(b). The intermediary should review the complaint against those categories and remove or restrict access where the rule applies.
Electronic impersonation is particularly relevant to modern synthetic media. Generative tools can reproduce a person’s face, voice, gestures, or apparent conduct. The new synthetic media provisions separately address deceptive depictions of natural persons and real-world events, while Rule 3(2)(b) provides the accelerated complaint pathway for the specified victim-focused categories.
For a content syndicator, this means “deepfake” should never be the only field used to determine a deadline. The case-management system should also record the affected person, type of depiction, complaint source, alleged harm, content URL or identifier, and the applicable rule.
Four Compliance Clocks Now Matter for Content Removal and Grievances
The 2026 rules create four response periods that content operations should keep separate: three hours, two hours, 36 hours, and seven days. The official FAQ published with the amendment lists these timelines together, which helps distinguish legal-order compliance from user grievance handling.
Three hours: Rule 3(1)(d) applies after actual knowledge through a competent court order or a reasoned written government intimation issued through the prescribed process. The intermediary must remove or restrict access within three hours.
Two hours: Rule 3(2)(b) applies to the specified categories involving private areas, nudity, sexual acts or conduct, or electronic impersonation, including artificially morphed images.
Thirty-six hours: Certain grievances seeking removal of information covered by specified parts of Rule 3(1)(b) must be resolved within 36 hours. This is a separate complaint route and should not be merged with the three-hour government or court-order process.
Seven days: The general period for resolving grievances under the grievance mechanism was reduced from 15 days to seven days. The requirement to acknowledge a grievance within 24 hours remains part of the grievance process.
For editorial and syndication teams, a single generic service-level target such as “remove AI content within three hours” is therefore too broad. A better operating model assigns the deadline only after the request has been legally and operationally classified.
The trigger also matters. A countdown should be connected to the recorded time of receipt of the relevant complaint, order, or government intimation. Teams should retain timestamped records so that the sequence of receipt, review, escalation, decision, removal, and restoration can later be reconstructed.
Content Syndicator Is an Industry Term, Not a Separate Legal Category Under the IT Rules
The IT Rules do not create a standalone legal category called a content syndicator. The duties that apply depend on what the business actually does. A syndication service can potentially perform functions associated with an intermediary, a publisher of news and current affairs content, a news aggregator, a news agency, or more than one role across different products.
The IT Rules separately define publishers of news and current affairs content. The definition includes online papers, news portals, news aggregators, news agencies, and similar entities.
Consider a syndication company that performs several activities:
- It receives third-party videos from creators.
- It stores those files in a content-management system.
- It automatically distributes them to partner sites.
- It operates its own editorial news portal.
- It aggregates stories from external publishers.
- It provides user-upload functionality.
Those activities should not automatically be treated as one regulatory function. The intermediary rules may apply to one part of the operation, while publisher obligations may apply to another.
This distinction also affects the takedown discussion. The three-hour and two-hour deadlines described above arise from intermediary due diligence and grievance provisions. A business should therefore establish which services fall within the intermediary framework before copying those deadlines across every publishing workflow.
For news syndicators, the publisher provisions can create additional duties beyond the synthetic media rules.
A useful compliance inventory should map each product, website, API, distribution service, upload feature, and editorial operation to the role it performs.
What Counts as Synthetically Generated Information Under the 2026 Rules
The 2026 rules define synthetically generated information, or SGI, as artificially or algorithmically created or modified audio, visual, or audiovisual content that appears real, authentic, or true and depicts a person or event in a manner that can be difficult to distinguish from reality. The definition is narrower than the broad everyday use of “AI-generated content.”
The rules therefore focus heavily on media that can imitate reality.
Examples that can fall within the concept include artificially generated or modified:
- Images of real people.
- Video depicting a person acting.
- Voice recordings that reproduce a person’s apparent speech.
- Audiovisual depictions of events presented as real.
- Manipulated media that materially changes what a person appears to have said or done.
The definition does not mean every use of generative technology becomes regulated synthetic media under the SGI provisions.
The official FAQ identifies exclusions for routine, good-faith editing and technical processes where the changes do not materially misrepresent the underlying information. Examples include correction, color adjustment, noise reduction, transcription, compression, accessibility work, translation, searchability improvements, and similar processing.
The FAQ also discusses good-faith creation of documents, presentations, PDFs, educational material, training material, and research material where the process does not create a false record or materially deceptive representation.
Text-only generative AI output is not treated as SGI under this specific synthetic media definition. Text can still be subject to the broader rules governing unlawful information, but the fact that text was produced by an AI model does not by itself place it inside the SGI definition.
This difference should appear in internal moderation taxonomies. A system that labels every AI-assisted article, headline, transcription, translated sentence, photo correction, generated image, synthetic voice, and deepfake with the same regulatory classification will create unnecessary confusion.
Lawful Synthetic Media Can Still Require Clear Labels and Provenance Information
The rules do not treat all synthetic media as prohibited. Where synthetically generated information is allowed to remain available, the intermediary duties include prominent labeling, audio disclosure where applicable, and permanent metadata or another provenance mechanism to the extent technically feasible.
For visual SGI, the label must be displayed prominently so that it is easily noticeable and adequately perceivable.
For audio SGI, the rules provide for a prominently prefixed audio disclosure.
The metadata requirement is also significant for syndication chains. The rules contemplate permanent metadata or another provenance mechanism, to the extent technically feasible, including a unique identifier that can help associate the synthetic material with its origin.
The technical-feasibility qualifier should remain part of any accurate description of the requirement. It is not accurate to rewrite the rule as an absolute requirement that every possible file format must contain identical metadata regardless of technical capability.
Intermediaries are also required not to enable users to modify, suppress, or remove the required label, metadata, or provenance mechanism.
That requirement has direct implications for content syndication.
A media asset may pass through an upload service, transcoder, content-management system, CDN, partner feed, social publishing tool, compression service, and downstream publisher. Each processing step can potentially strip metadata or alter visible disclosures.
A syndication workflow therefore needs to test whether:
- Image optimization preserves required provenance data.
- Video transcoding retains relevant identifiers.
- Audio conversion preserves the required disclosure.
- Cropping or resizing leaves visual labels adequately perceivable.
- Partner exports retain the required metadata.
- CMS previews do not unintentionally remove disclosures.
- Download and re-upload processes preserve the information required by the rules.
The legal requirement and the technical distribution chain need to be examined together.
Significant Social Media Intermediaries Have Additional Pre-Publication Duties
A Significant Social Media Intermediary, or SSMI, has additional duties for detecting and identifying synthetic media before publication. India currently uses a threshold of 50 lakh, or five million, registered users for classification as an SSMI.
Under the 2026 synthetic media provisions, an SSMI must require a user to declare whether information being submitted is synthetically generated information before it is displayed, uploaded, published, transmitted, or shared through the service.
The SSMI must also use technical measures to verify the declaration. Where the content is identified as synthetic, the applicable labeling requirements follow.
This requirement should not be rewritten as a universal declaration duty applying identically to every website, publisher, small intermediary, or content syndicator in India. The specific pre-publication declaration and verification provision applies to SSMIs.
A syndication business that operates social features should therefore determine whether those services meet the applicable SSMI definition and user threshold.
Technical verification can also affect product design. Upload systems may need fields recording the user’s declaration, machine-detection output, review status, label status, provenance status, and any later correction.
These records can also help when the same file reaches multiple versions or URLs through a syndication network.
Synthetic Media Prohibitions Cover More Than Intimate Deepfakes
The rules identify several categories of synthetic material that intermediaries must prohibit through their rules, regulations, privacy policy, and user agreement. The covered categories extend beyond sexual deepfakes and include deceptive depictions capable of presenting a natural person or real-world event falsely as authentic.
The notified provisions address synthetic information connected with categories such as child sexual exploitation or abuse material, non-consensual intimate imagery, obscene or sexually explicit material, bodily privacy violations, certain false electronic records, and specified prohibited material connected with explosives, arms, or ammunition.
The rules also address deceptive false depictions involving the identity, voice, conduct, action, statement, or event associated with a natural person or real-world occurrence.
For syndicators, that means moderation cannot rely only on an “adult deepfake” category.
The classification system may need separate fields for:
- Identity impersonation.
- Voice cloning.
- Fabricated conduct.
- Fabricated statements.
- Altered real-world events.
- Non-consensual intimate material.
- Child sexual exploitation or abuse material.
- Obscene or sexually explicit synthetic content.
- False records generated through synthetic processes.
The exact legal treatment still depends on the rule involved, applicable law, the type of notice received, and the intermediary’s role.
A Content Syndicator Needs a Deadline-Based Takedown Workflow
A practical compliance workflow should begin by classifying the notice before assigning the deadline. The source of the request, legal basis, content type, affected person, and platform role determine whether the three-hour, two-hour, 36-hour, or seven-day process applies.
A workable internal sequence can follow these stages.
Record receipt immediately. Store the received time, sender, delivery channel, URL or content identifier, account information where relevant, affected person, attached materials, and initial content category.
Classify the trigger. Determine whether the item is a court order, reasoned government intimation, Rule 3(2)(b) victim complaint, another specified removal grievance, or a general grievance.
Assign the correct clock. Do not rely on a generic “deepfake deadline.” The response timer should be tied to the applicable rule.
Confirm the content location. Syndicated material can exist at an original URL, cached URL, partner endpoint, API response, thumbnail URL, video rendition, embedded player, or downloadable asset. The service should identify the versions it actually controls.
Escalate short-deadline cases. The two-hour and three-hour periods require direct routing to personnel who can assess and execute access restrictions without the case sitting in a normal customer-support queue.
Execute the required action. Depending on the applicable rule and legal instruction, the intermediary may need to remove the information or restrict access to it.
Record completion. Keep timestamps for receipt, review, escalation, decision, action, and any communication sent through the grievance process.
Check syndicated copies under the service’s control. Removal from a primary article page does not necessarily remove a video rendition, cached copy, partner feed, thumbnail, or content object stored elsewhere in the same system.
Preserve required records. A business should retain the material and operational records required by applicable law and its legal retention process even when public access is disabled.
Safe Harbor Now Depends Heavily on Documented Due Diligence
Section 79 of the Information Technology Act provides conditional protection for qualifying intermediaries in relation to third-party information. Rule 7 of the IT Rules states that where an intermediary fails to observe the rules, the exemption under Section 79(1) will not apply and the intermediary can be subject to liability under applicable law.
The relationship should be described carefully. A missed internal target and a final legal determination about Section 79 are not identical concepts. The rules establish due-diligence conditions whose non-observance can make the statutory exemption unavailable.
The 2026 official FAQ also addresses the opposite concern. It explains that removal of synthetic information in compliance with the rules, including use of the prescribed technical measures, does not by itself violate the conditions in Section 79(2)(a) or Section 79(2)(b).
For syndication services, recordkeeping is therefore part of the compliance process.
Useful records can include:
- The original notice or complaint.
- Timestamp of receipt.
- Legal or grievance category assigned.
- URL and content identifier.
- Copy of the relevant content.
- SGI classification.
- User declaration where required.
- Technical verification result where required.
- Label and provenance status.
- Escalation timestamps.
- Removal or access-disablement timestamp.
- Restoration information if later applicable.
- Communications with the complainant or issuing authority.
The goal is to make the operational history of each case clear.
User Notices Must Also Reflect the Synthetic Media Rules
Intermediary compliance is not limited to content removal. The amended rules require intermediaries to periodically inform users about their rules and regulations, privacy policy, user agreement, prohibited information, and consequences of non-compliance, with the notification occurring at least once every three months.
For a service handling synthetic media, user-facing policies should accurately explain the categories of SGI that users may not create, upload, publish, transmit, share, or disseminate through the service where the rule applies.
Policies should also match the actual product.
If the upload screen requests an SGI declaration, the terms should explain that process.
If synthetic audio receives a disclosure, the publishing policy should describe how it appears.
If metadata is retained through an asset pipeline, internal documentation should explain which formats and processing stages support it.
If user accounts can be suspended or access can be restricted for prohibited use, the consequences described in the user agreement should correspond with the operational system.
A policy that describes controls the product does not actually implement creates avoidable compliance gaps.
Common Misreadings of India’s 2026 Synthetic Media Rules
Several interpretations can lead syndicators to build the wrong workflow. The notified text and official FAQ provide a more precise reading.
“Every deepfake must be removed within two hours” is too broad. The two-hour rule applies to the specified Rule 3(2)(b) complaint categories, including electronic impersonation and artificially morphed imagery in that context. Other synthetic media can enter different processes.
“Every complaint starts the three-hour timer” is incorrect. The three-hour Rule 3(1)(d) period is tied to actual knowledge received through a competent court order or a reasoned written intimation from the appropriate government or authorized agency.
“All AI-generated information is SGI” is incorrect. The specific SGI definition covers audio, visual, and audiovisual material meeting the definition. The official FAQ says text-only AI output is outside this SGI definition, though other IT Rules can still apply.
“Every AI-assisted edit needs a synthetic media label” is too broad. Routine good-faith corrections and specified technical or accessibility processes are excluded where they do not materially misrepresent the information.
“Every intermediary must ask every uploader for an SGI declaration” overstates the provision. The express pre-publication declaration and technical verification duty applies to Significant Social Media Intermediaries.
“Provenance metadata is required in every technical circumstance” omits an important qualifier. The notified text uses the standard “to the extent technically feasible” for permanent metadata or another provenance mechanism.
“Content syndicator is one defined category with one compliance checklist” is inaccurate. Legal duties depend on whether the business is acting as an intermediary, publisher, news aggregator, news agency, SSMI, or through another covered function.
What Indian Content Syndicators Should Change in Their Compliance Systems
Indian content syndicators should treat the 2026 amendments as a routing, classification, metadata, labeling, and response-time problem, not merely as a faster moderation policy. A compliant process needs to know what content is being distributed, what role the service performs, how a request entered the system, and which legal deadline applies.
The highest-priority operational changes are straightforward.
- Create separate intake categories for court orders, government intimations, affected-person complaints, other removal grievances, and general grievances.
- Configure three-hour, two-hour, 36-hour, and seven-day workflows separately.
- Preserve the 24-hour acknowledgment step for the general grievance process.
- Record exact receipt and action timestamps.
- Add a clear SGI classification to media workflows.
- Distinguish synthetic images, audio, and video from text-only AI output and routine excluded edits.
- Verify that required labels survive resizing, transcoding, compression, export, caching, and downstream distribution.
- Preserve provenance metadata where the rule requires it and where technically feasible.
- Prevent product features from intentionally removing required labels or provenance information.
- Determine whether any service qualifies as an SSMI and, if so, implement user declarations and technical verification before publication.
- Review user terms and recurring notices so synthetic media restrictions are accurately described.
- Map every syndication product by legal function rather than applying one label to the entire company.
- Maintain an escalation path that can execute short-deadline removal or access restrictions outside ordinary editorial queues.
The central compliance lesson is precision. Three hours, two hours, 36 hours, and seven days are different legal clocks with different triggers. A system that correctly classifies those triggers is more useful than a blanket policy that treats every AI-related complaint as the same event.
As of September 2026, the February 2026 notified amendments remain the primary official basis for these synthetic media obligations. Content operations should rely on the Gazette, the consolidated IT Rules, and current government guidance when updating internal procedures, particularly where online summaries still mix draft proposals with provisions that have already taken effect.
India’s 2026 IT Rules make synthetic media compliance a time-sensitive operational responsibility for intermediaries and content distribution services. The key requirement is not simply removing AI-generated content quickly. Platforms must first identify the type of complaint or legal notice, determine the applicable rule, and then follow the correct deadline, whether that is two hours, three hours, 36 hours, or seven days.
For content syndicators, compliance also extends beyond takedowns. Synthetic media classification, visible disclosures, provenance metadata, user declarations for qualifying Significant Social Media Intermediaries, grievance handling, recordkeeping, and preservation of labels across distribution systems all require attention. Businesses that map their legal role clearly and build deadline-based workflows around the notified rules will be better prepared to respond accurately when synthetic media complaints, court orders, or government directions arrive.
New IT Rules on Synthetic Media: FAQs
India’s 2026 IT Rules introduce stricter obligations for intermediaries handling synthetically generated information, including faster takedown timelines, labeling requirements, provenance measures, grievance handling, and additional duties for Significant Social Media Intermediaries.
The three-hour deadline applies when an intermediary receives actual knowledge through a competent court order or a reasoned written intimation from the appropriate government or an authorized agency requiring removal or disabling of unlawful information.
No. The two-hour deadline applies to specified complaints involving private areas, nudity, sexual acts or conduct, and electronic impersonation, including certain artificially morphed images. Other forms of synthetic media can fall under different procedures and timelines.
The key timelines include three hours for qualifying court orders or government intimations, two hours for specified victim complaints, 36 hours for certain other removal grievances, and seven days for general grievance resolution. General grievances must also be acknowledged within 24 hours.
Synthetically generated information generally refers to artificially or algorithmically created or modified audio, visual, or audiovisual material that appears authentic and depicts a person or event in a way that can be difficult to distinguish from reality.
Text-only AI output is not treated as synthetically generated information under the specific synthetic media definition. However, AI-generated text can still be subject to other provisions of the IT Rules and applicable Indian laws.
Where the relevant synthetic media provisions apply, intermediaries may need to provide prominent visual labels, audio disclosures, and provenance information. Permanent metadata or another provenance mechanism is required to the extent technically feasible.
No. The specific requirement for users to declare whether submitted content is synthetically generated applies to Significant Social Media Intermediaries. These intermediaries must also use technical measures to verify the declaration.
A Significant Social Media Intermediary is a social media intermediary that meets the prescribed threshold of registered users in India. The current threshold is 50 lakh, or five million, registered users in India.
Content syndicators should classify complaints by legal trigger, maintain separate workflows for different deadlines, preserve required labels and provenance information, record timestamps, review user policies, test metadata retention across distribution systems, and determine whether each service operates as an intermediary, publisher, news aggregator, or another regulated category.