On 3 November 2026, X told SFLC.in and the Internet Freedom Foundation (IFF) that the Delhi Police had asked it to remove 42 URLs, including their own accounts and posts documenting censorship. Days later, X told the Delhi High Court that its internal investigation found the notices came from someone impersonating the Delhi Police. Justice Amit Mahajan recorded that statement and disposed of the petitions. The case ended quickly. The structural problem it exposed did not.
What happened
According to MediaNama's report of the hearing, X's counsel said the platform took no action on the request, called it illegal, and said it had filed a complaint with the Delhi Police against the impostor. Senior Advocate Akhil Sibal, appearing for both organisations, said the notices named the Special Commissioner of Police (Crime) as sender. SFLC.in had emailed the address on the notice to ask whether it was genuine, and the email bounced. A Delhi Police advocate attended but made no submissions. The court gave the organisations liberty to return if any action is taken.
The context matters. MediaNama's account of the petition says the later notice sought removal of 37 posts and withholding of three accounts, invoked Section 356 of the Bharatiya Nyaya Sanhita (criminal defamation), Section 79(3)(b) of the IT Act and Rule 3(1)(d) of the IT Rules 2021, and demanded action within three hours. EFF reports the grounds cited included hate speech, alleged defamation of the RSS and security concerns, and that the targets included the Free Speech Tracker and the account @geoblockedindia, which publishes details of content blocked in India.
No one has publicly identified the impostor. X's finding is X's own, and the Delhi Police made no submission on it.
The strongest case for fast takedown channels
There is a serious argument for the system that makes this episode possible. Platforms receive enormous volumes of harmful content, from incitement to non-consensual imagery, where hours matter. Police need a direct line to intermediaries, and requiring a court order for every removal would leave victims waiting. Rule 3(1)(d) of the IT Rules gives platforms a clear trigger for acting, and in September 2025 the Karnataka High Court, per LiveLaw's report, dismissed X Corp's challenge to the Sahyog portal. Justice M. Nagaprasanna held that the portal was conceived under Section 79(3)(b) and Rule 3(1)(d) and rejected the label "censorship portal." Speed and accessibility are legitimate goals.
Why speed without verification fails
The incident shows what happens when speed is the only design value. Three weaknesses stand out.
- Authentication is weak. A notice that arrives as an email-style report, with a sender title and a deadline measured in hours, can be forged. X caught this one. A smaller platform with a thinner legal team may not, and a platform that complies by default would have removed the accounts of two civil-society groups on the strength of a forgery.
- Targets learn about it by accident. Under India's formal blocking regime, Rule 8 of the 2009 Blocking Rules requires notice to the hosting party and allows no fewer than 48 hours to respond. Rule 16 requires "strict confidentiality" about requests and actions. The Supreme Court upheld that regime in Shreya Singhal v. Union of India (2015), in part because it requires written reasons and a hearing process. Informal police-to-platform requests carry none of those safeguards, and X's decision to notify the affected organisations was a courtesy, not a legal duty. The petitioners learned of the notices because X chose transparency.
- Documenting censorship became grounds for censorship. The notices objected to the tracking of blocking orders itself. A free-speech watchdog's work product, a record of what was restricted and when, was treated as the offending content. That is a proportionality failure whoever sent the notices. Whether genuine or forged, the request treated documentation of state action as unlawful speech.
There is also a verification gap that cuts the other way. If forged notices can look credible enough to prompt an X internal investigation and a court appearance, genuine ones become harder to trust. Platforms are pushed toward either over-compliance or blanket suspicion. Neither serves users.
What proportionate reform looks like
None of this requires abandoning rapid takedown channels. It requires making them auditable.
- Authenticated submission. Government requests should arrive through a single verified channel with signed credentials. A platform should be able to confirm in seconds that an order is real, and a notice from an unverified email should carry no legal weight.
- Disclosure to affected users by default. Where there is no genuine investigative need for secrecy, the account holder should be told which authority asked, under which provision, and how to challenge it. Rule 16 confidentiality should not become a shield for requests that never passed through the formal process.
- Published statistics. Platforms and authorities should report volumes, legal bases and outcomes, so that patterns are visible. Documentation projects like SFLC.in's tracker exist because the state does not publish this data itself.
- Protection for documentation. Records of blocking orders and the analysis of them should be treated as legitimate speech about matters of public concern, not as grounds for removal.
The pro-innovation case for these changes is practical. Intermediaries cannot build compliance systems around channels they cannot authenticate, and uncertainty falls hardest on smaller services. Clear, verifiable procedures reduce legal risk for platforms and preserve public trust in the legitimate use of takedown powers.
The bottom line
The Delhi High Court disposed of these petitions after a single hearing, and X deserves credit for refusing the request and telling the targets. But a system that depends on a platform's discretion to catch forgeries is fragile. Until takedown requests can be authenticated, disclosed and reviewed, the next forged or overbroad notice will test whether the platform in its path is as careful as X was this time.