Policymakers in New Delhi are weighing a separate legal definition for an AI intermediary, a question that decides whether chatbot makers keep the safe harbour protection granted to social media platforms. Mint reported the debate on 28 October, saying lawyers and Big Tech firms have made representations to the Ministry of Electronics and Information Technology, known as MeitY, and submitted policy whitepapers.
At issue is how the law treats services that generate answers rather than carry posts written by users. Social media platforms receive safe harbour protection under the Information Technology Act, 2000 and the 2021 intermediary guidelines, which shield platforms from prosecution over content posted through them. Chatbots work differently from an intermediary carrying user posts: services such as OpenAI’s ChatGPT, Google’s Gemini, Anthropic’s Claude and Meta’s Meta AI generate results from their own data and judgement in response to user prompts. That difference complicates the existing framework, and it is why a new intermediary definition is under discussion.
The discussion comes six months after the Centre proposed an amendment to regulate AI-generated content on social media. On 8 October, IT minister Ashwini Vaishnaw said his ministry would release a consultation paper for a dedicated AI law. That paper has not yet appeared, and the shape of any law remains open, including whether India adopts a European Union style AI statute or sector by sector rules.
What an AI intermediary definition would change
A definition would draw a line between AI platforms and the online platforms already covered by law. Lawyers who have made representations to MeitY say a dedicated statute would be a step in the right direction, according to Mint. Their argument is that AI systems need rules written for what they do, rather than borrowed from social media.
Big Tech executives take a different view. They have pushed for sector-wise AI rules rather than an overarching law, Mint reported. Their position is that AI is used in different ways across finance, health care, education and media, and that rules should follow the use rather than the technology. A single statute, in that view, risks covering unlike cases with one set of duties.
The disagreement also covers liability. A senior executive at a leading Big Tech company, speaking anonymously to Mint because the conversations are private, said users and platforms share liability for generated content, and that court verdicts abroad have found AI platforms editorialise to some degree. That matters because editorial control is one of the tests courts use when deciding whether a platform can claim the protection of a mere carrier.
Why intermediary status matters for chatbots
Safe harbour is the protection that lets a platform operate without prosecution for every item posted through it. If an AI intermediary keeps that protection, its legal exposure for generated output stays limited, provided it meets the conditions the law sets. If it loses that protection, or if a new definition places it outside the intermediary category, it could face direct liability for what its system produces.
That exposure would land on companies with very large user bases in India. OpenAI has said India has more than 100 million weekly active users, making it one of the largest markets for AI in the world. The stakes rose over the past year after a series of grey area content incidents on AI platforms, including sexualised images generated through Grok, the chatbot on Elon Musk’s X, which Elets CIO cited.
An executive at a second firm told Mint that no standalone talks on a new AI intermediary definition have taken place, though the subject has featured in ongoing discussions with MeitY about AI-generated content on social media. That account suggests the definition question is moving inside a wider conversation about online content, rather than a separate negotiation.
What happens next for intermediary rules
The next public step is the consultation paper Vaishnaw promised on 8 October. MeitY secretary S Krishnan has said existing legal provisions have handled early concerns about deepfakes and AI-generated synthetic content, but that an additional regulation or law may be needed as the technology spreads. Vaishnaw has also said the current information technology law was framed long before AI emerged.
India still has no unified AI law. Most AI services fall under the Information Technology Act, 2000 and its rules, alongside the Digital Personal Data Protection Act, 2023 and, where relevant, the Copyright Act, 1957. That patchwork applies depending on the activity or harm involved.
Whether the consultation paper proposes an intermediary definition, and whether it leans toward one statute or sector-wise rules, will show which side of the current argument has carried more weight with MeitY.