India's Ministry of Electronics and Information Technology is preparing a consultation paper that will spell out how social platforms must label synthetic media and remove deepfakes that impersonate public officials, building on IT Rules amendments that took force in February 2026.
Minister Ashwini Vaishnaw told reporters the paper should arrive within a month, with deepfake harms, user safety and cybersecurity among the core chapters. The exercise stops short of a standalone AI Act for now, leaning instead on the IT Act, the Digital Personal Data Protection Act and sector regulators.
What platforms already face
Amendments notified on 10 February define synthetically generated information and require significant social media intermediaries to label AI-created audio and video, preserve metadata and respond faster to government-flagged posts that could disturb public order. An 8 October advisory went further, asking networks to contextualise manipulated or out-of-context clips presented as current events.
MediaNama reported that the advisory tells platforms to prioritise complaints from authorised government agencies when content has law-and-order implications. That line matters during election seasons and festival weeks, when archived protest footage often recirculates as if it were live.
Why deepfakes moved to the front
Vaishnaw has repeatedly cited synthetic videos of ministers and film stars as a catalyst. Telugu industry figures protested a viral deepfake in early October, and the Home Ministry's cybercrime unit later warned that similar clips were being stitched into fake welfare advertisements on Meta properties.
The consultation is expected to ask whether voluntary commitments from large language model vendors are enough, or whether MeitY needs binding duties on dataset documentation, red-team testing and incident reporting. India's November 2025 AI governance guidelines had promised to revisit mandatory rules after nine to twelve months if self-regulation failed.
Implications for enterprises
Banks, insurers and hospitals already deploy generative tools for customer service. The paper will likely ask how those firms should log training data, especially when models ingest Indian-language speech. CDAC and the IndiaAI mission are expanding GPU capacity, so policymakers want guardrails before more state data lands in shared clusters.
Developers should expect questions on watermarking standards compatible with open-source weights, and on liability when a chatbot gives harmful advice. None of that replaces sector rules — the Reserve Bank of India and the Telecom Regulatory Authority of India retain their own channels — but it sets a baseline for anything touching consumer-facing apps.
How to participate
Once published, the consultation will run on MeitY's portal with a comment window, typically 30 to 45 days. Civil society groups that challenged the October labelling advisory on free-expression grounds are likely to file counter-comments urging narrower government priority queues.
Until final rules land, firms should map their content moderation pipelines to the February SGI obligations: user-facing labels, grievance officer escalations and takedown clocks. The winter paper will not erase those duties; it will decide how much harder they become.