India's Information Technology Amendment Rules took effect on 20 February 2026, imposing duties on intermediaries that enable the creation or circulation of covered synthetic media. They require technical measures against unlawful synthetic content and labelling of permissible material. For international platforms, the legislation reaches product design and publication workflows.1
The rules require visible labels for visual content and an audio disclosure before audio content. They also require permanent metadata or other appropriate provenance mechanisms, including an identifier for the intermediary's computer resource, to the extent technically feasible. Significant social media intermediaries must obtain user declarations about synthetic content, take technical steps to verify them and label confirmed synthetic material.1
The definition leaves some tools outside
The Ministry of Electronics and Information Technology's explanatory questions distinguish realistic synthetic audio, images and video from text-only outputs. A chatbot's written answer is not itself synthetically generated information under this definition, although ordinary duties concerning unlawful information still apply. The distinction matters for providers selling several kinds of AI service through the same interface.2
The ministry also identifies exclusions for routine, good-faith editing and accessibility improvements that do not materially change the underlying content. Its examples include reducing background noise, adding captions and preparing illustrative training material. These are qualifications to the definition, rather than exemptions for an entire company or a general permission to create deceptive records.2
In a March parliamentary explanation, the government described the aim as helping users identify artificial material and preventing deception or misuse. Its Hindi statement uses the phrase:3
स्पष्ट लेबलिंग और पता लगाने योग्य मेटाडेटा
Clear labelling and traceable metadata.3
That explanation places the rules within a wider programme of user warnings, action against impersonation and protection against harmful synthetic images. It presents identification of artificial content as a safeguard for users, rather than a prohibition on all legitimate AI-assisted production.3
Global services dispute the implementation burden
The Computer & Communications Industry Association argues that the obligations can require service-specific redesign, additional compliance systems and continuing monitoring across globally integrated platforms. It warns that upstream providers may lack visibility into user outputs or the ability to apply content-level controls. These are the trade association's objections, not measured evidence of a fall in India's digital-services imports.4
CCIA also questions whether imperfect detection, combined with faster removal requirements, will encourage excessive takedowns. Its request to reopen consultation and pause implementation is an industry recommendation; it does not itself suspend the notified rules.4
The Internet Freedom Foundation welcomes exclusions for ordinary editing but remains concerned about compulsory labelling, unreliable detection and restrictions before publication. It argues that deciding whether editing is in good faith will often fall to intermediaries. The unresolved operational question is how platforms will make that distinction consistently while preserving legitimate creative and accessibility uses.5