MeitY's Actions Against X over Grok (March 2025 and January 2026)

March 2025. The first Indian engagement with Grok concerned language, not images. In March 2025 users noticed that Grok, X's chatbot, responded to prompts in Hindi slang and with abusive language. The IT Ministry told the Press Trust of India that it had reached out to X and would examine what led to the responses; MediaNama reported no formal notice and no outcome. MediaNama's founder Nikhil Pahwa argued that the problem lay with the input ('garbage in, garbage out') and that Grok, trained on X, mirrors the tone of discourse on the platform.¹

December 2025. In late December 2025 users began asking Grok, in public replies on X, to alter photographs of real people, mostly women, by removing clothing or adding suggestive poses, and the edited images appeared in replies on the platform.⁴⁶ MediaNama reported that on 1 January 2026 Grok acknowledged having generated an AI image of two young girls estimated to be between 12 and 16 years old.⁴ The feature had launched around Christmas 2025. On 29 December 2025 MeitY's Cyber Laws Division had separately sent an advisory to intermediaries, including social media platforms, to curb 'obscene, pornographic, vulgar, and other unlawful content'; the advisory did not mention AI-generated content or deepfakes. It asked intermediaries to review their compliance frameworks, moderation practices and user enforcement, noting that significant social media intermediaries must deploy automated tools to prevent unlawful content, and warning of the loss of Section 79 protection and of prosecution under Sections 67, 67A and 67B of the IT Act, the Bharatiya Nyaya Sanhita and the Indecent Representation of Women (Prohibition) Act, 1986.⁸

On 2 January 2026 MeitY wrote to X Corp. According to MediaNama and the Internet Freedom Foundation, the letter cited X's failure to meet due-diligence obligations under the IT Act, 2000 and the IT Rules, 2021, and required X to:³¹⁰

• take immediate and concrete steps to prevent the hosting, generation or sharing of obscene, nude, sexually explicit or otherwise unlawful content, including content made through Grok and other xAI tools;
• carry out a 'comprehensive technical, procedural and governance-level review' of the Grok application, covering prompt processing, output generation, image handling and safety guardrails;
• enforce X's user terms and AI usage rules, including suspension or termination of violating accounts; and
• submit a detailed Action Taken Report (ATR) within 72 hours covering measures taken or proposed, the Chief Compliance Officer's role and oversight, action against content and accounts, and mechanisms for mandatory reporting of offences under criminal procedure law.

MeitY warned that non-compliance could lead to serious legal consequences, including loss of safe harbour. MediaNama's reporting noted that if Grok itself generated explicit imagery, X could be treated as the publisher of illegal content rather than a neutral platform unless it showed due diligence.⁶ In an interview MeitY Secretary S. Krishnan said that platforms 'cannot escape their duty or responsibility simply by pleading safe harbour'.³ Economic Times reports, cited by MediaNama, quoted unnamed officials as saying that the government was prepared to revoke X's safe harbour if it did not comply with the latest takedown directions.⁴

X responded in stages.⁴⁵⁶⁷⁹

• 4 January 2026: X's Safety account and Elon Musk posted that users who prompted Grok to make illegal content, including child sexual abuse material, would face the same consequences as if they had uploaded it.
• 5 to 7 January: The deadline for the ATR was first 5 January; X asked for time and MeitY extended it by 48 hours to 7 January. X sent a detailed reply after the extension that acknowledged Indian law and described its takedown policies for non-consensual intimate images and misleading posts.
• 8 January: Citing government sources, PTI and MediaNama reported that MeitY found the ATR inadequate because it lacked specifics on takedown actions, on the specific action taken against Grok AI and on how X would prevent recurrence, and sought further information, in particular on takedowns of Grok-generated obscene content.
• 9 January: X limited Grok's image generation and editing on X to paying subscribers, though the standalone Grok website still allowed image generation.
• Around mid-January: Indian government sources said X acknowledged moderation lapses, removed about 3,500 items of content and deleted more than 600 accounts. MediaNama noted that the figures came from unnamed government sources and had no timeframe.
• 15 January: X said it had added technological measures to stop the Grok account from editing images of real people into revealing clothing, applying to all users, and updated its geoblocking to block the capability where it is illegal. MediaNama's test of the standalone Grok website found that it still complied with many such requests.

MediaNama reported on 4 February 2026 that Grok on X was still editing images of real people to remove clothing, apparently only for verified premium users.²

Is Grok's output third-party content? Section 79 protects an intermediary only for third-party information. MediaNama's founder Nikhil Pahwa argued, in a piece MediaNama cited, that X publishes Grok's outputs through its own AI service and is therefore not acting as an intermediary for them, so safe harbour should not apply, and that X's terms shift responsibility to users who prompt Grok while X itself may be liable for generating non-consensual intimate imagery.⁴ The report presented only that side. In the reports reviewed, MeitY's notice did not resolve the question; it framed the matter as a failure of due diligence under the IT Act and the IT Rules.³¹⁰

Gap in the framework. The Internet Freedom Foundation, writing in January 2026, said MeitY had not opened a dedicated process for Grok as an AI system and had relied on takedowns and platform enforcement under the intermediary liability regime. It said that a standalone chatbot that is not an intermediary could fall into a 'regulatory grey zone', that India lacked requirements for pre-deployment testing, safety-by-design and independent oversight of high-risk generative AI, and that the November 2025 AI Governance Guidelines deferred to existing law. It cited the IT Act's Sections 66E, 67, 67A, 67B and 85 (liability of companies), the BNS, the Indecent Representation of Women (Prohibition) Act, the POCSO Act and the Young Persons (Harmful Publications) Act as the relevant offences, and asked for a consent-based rather than obscenity-based treatment of non-consensual intimate imagery and for a white paper on generative AI safety within 30 days.⁹¹⁰

Opacity. The advisory of 29 December 2025 was obtained by the Internet Freedom Foundation through a right-to-information request and had not been posted on the MeitY website when IFF published.⁹ In its submission on the March 2026 draft amendments to the IT Rules, IFF described the 2 January letter to X as a privately issued MeitY letter and cited it as an example of an executive instrument that is not public yet is used to pressure platforms.¹²

The Grok episode is the clearest instance so far of the Union government using the safe-harbour regime directly against a generative AI feature. Three features stand out. First, MeitY acted within days, through a letter with a deadline, rather than through legislation or a formal order, using the same lever as in the 2023 deepfake advisories: a threat to Section 79 protection.³⁸ Second, it asked for a systemic review of the AI product (prompt processing, output generation, image handling and guardrails), which goes beyond content removal and resembles a safety audit, although the IT Rules in force at the time contained no express requirement for such a review. Third, the notice was addressed to X, the platform that hosts the chatbot; it refers to xAI's tools, but the reports reviewed do not describe any direction to xAI itself.

The case also fed into the rule changes that followed. The IT Rules amendment that came into force on 20 February 2026 requires any intermediary that offers a computer resource capable of generating synthetic media to deploy technical measures to prevent non-consensual intimate imagery, child sexual abuse material and other unlawful synthetic content, and MeitY's FAQs list synthetic 'undressing' content among examples of prohibited synthetic imagery.¹¹

As of 15 October 2026 the sources reviewed report no penalty, adjudication, revocation of safe harbour or court proceeding arising from the January 2026 notice, and no statement from MeitY closing the matter. X had by then restricted Grok's image tool as described above, although MediaNama reported in February 2026 that Grok on X could still edit images of real people in some circumstances.² The specific provisions on synthetic content that now govern such conduct (Rule 3(3) and Rule 4(1A) of the IT Rules) are described in the entry on the IT Rules amendment.