London, 2nd Nov 2026: India now gives platforms three hours to pull content the government flags. The aim, stopping deepfakes and fraud, is fair. But journalism, satire and political criticism can be caught in the same net, and few newsrooms are reporting it that way.
In February, India gave social media platforms three hours to remove content once a court or an authorised government officer flags it as unlawful. The earlier deadline was 36 hours. For non-consensual intimate images, including deepfake nudes, the window is now two hours, down from 24.
The goal is legitimate: curb deepfakes, impersonation, fraud and abuse. Most coverage stopped there and called it a “deepfake crackdown”. That is half the story. The other half is what happens to journalism, satire and dissent when a platform has 180 minutes to decide.
A country primed for fakes
The World Economic Forum’s Global Risks Report 2024 ranked India first in the world for misinformation and disinformation risk. The US was sixth and the UK eleventh. In the 2025 report, Indian executives still listed it among the country’s top five risks.
The money trail is real. Indians lost about ₹19,813 crore to cyber fraud in 2025 across 21.77 lakh complaints, according to the Indian Cyber Crime Coordination Centre (I4C). That covers all cyber fraud, not deepfakes alone. AI makes the con easier: a McAfee survey found 47% of Indian adults had faced an AI voice scam or knew someone who had, almost double the global average of 25%. Of Indian victims, 83% lost money.
Fact-checkers see the shift. Alt News fact-checked seven AI-generated images or videos in 2024. In 2025 it checked 39.
But the 2024 general election is a warning against hype. BOOM studied 258 election fact-checks from March to May 2024. Only 12 involved AI, and nine of those were cheap voice clones. Old videos, misleading edits and communal claims did most of the damage. The bigger AI story was legal and paid for: parties spent an estimated $50 million on authorised AI campaign content, and more than 50 million AI voice-clone calls went out in the two months before voting began, a messaging operator told WIRED. A law aimed only at deepfakes misses most of the problem.
Added after the consultation
MeitY notified the IT Amendment Rules on 10 February 2026. They took effect ten days later. The rules define “synthetically generated information”, require visible labels and embedded metadata on AI content, and compress every takedown deadline.
The three-hour clause was not in the draft MeitY released for comment on 22 October 2025. It appeared only in the final rules. In March, MediaNama filed an RTI application asking whether anyone was consulted on it.
That matters. A rule that changes how fast speech can be removed should face public scrutiny before it becomes law. With three hours on the clock, the safe choice for a platform is to remove first and check later.
Where satire falls
The rules exclude routine editing such as colour correction or compression, the preparation of documents, educational and research material, and accessibility features. Satire, parody and mimicry are not named.
Law firm Nishith Desai Associates notes that obvious parody may escape, but satire, re-enactments and embellished content sit in a grey area, because whether something “appears real” varies from viewer to viewer. In its feedback on the draft, JNU’s Centre for the Study of Law and Governance asked MeitY to carve out satire, parody and mimicry. The final rules did not.
Think of a realistic AI sketch of a minister, a reconstructed scene in an investigative documentary, or a source’s voice altered to protect them. Each could be flagged as synthetic and pulled before any editor is asked.
Takedowns already run at scale
The removal machine predates the AI rules. The Home Ministry’s annual report shows more than one lakh posts blocked in 2024-25, about 290 a day. Through the government’s Sahyog portal, over 2,300 blocking orders went to 19 platforms between October 2024 and October 2025, the Indian Express reported from RTI data.
Sahyog orders get no independent review, and users are often not told who wanted their post removed. Journalist Ila Kazmi filmed a police facial-recognition van at Jantar Mantar. Her reel was withheld in India in July and her whole Instagram account in September. No agency, officer or reason was given. On 9 October the Delhi High Court ordered the government and Meta to produce the takedown orders. The case returns on 17 November. How do you challenge an order when you cannot see who issued it, or why?
The courts have not settled the bigger question. The Karnataka High Court upheld Sahyog in September 2025, rejecting X Corp’s challenge. Comedian Kunal Kamra and senior advocate Haresh Jagtiani challenged Sahyog and the takedown rule in the Bombay High Court in February 2026. On 22 July the Supreme Court stayed those High Court hearings while it weighs the Centre’s plea to club the cases.
Meanwhile India fell to 157th of 180 countries in RSF’s 2026 World Press Freedom Index, down from 151st. The Reuters Institute’s 2024 India report had already flagged fears that new laws could undermine press freedom.
Two bodies to watch
The Election Commission. In January 2025 the ECI asked parties to label AI campaign material “AI-Generated”, “Digitally Enhanced” or “Synthetic Content”. Before the Bihar polls in October 2025 it went further: labels must cover 10% of the display, and misleading content must come down within three hours of notice. These are advisories, not statute.
The AI Governance and Economic Group (AIGEG). MeitY constituted it on 13 April 2026 and announced it on 16 April, with IT Minister Ashwini Vaishnaw in the chair. Its mandate includes sorting AI uses into “deploy”, “pilot” and “defer”. MediaNama found it left out TRAI, the Competition Commission and the Data Protection Board, though the government’s own AI guidelines recommended them. It coordinates; it cannot enforce.
What comes next
On 8 October Vaishnaw said MeitY would publish an AI regulation consultation paper within a month, covering safety and deepfakes. Later that month he said it would come “in the coming weeks”. That paper should answer four questions:
- Transparency: will takedown orders be published, naming who issued them and why?
- Press freedom: will journalism, satire, reconstructions and source protection get a written exemption?
- Due process: can a creator get notice and appeal before content comes down, not after?
- Accountability: who checks the officers who press the button?
India’s AI work in farming, health and Indian-language tech is real and worth covering. So is the delete button being built beside it. Three hours makes content disappear faster. It does not make the decision to remove it any more accurate.
Submit your review | |