Training data risk, IP ownership over AI-assisted output, model licensing, and the enterprise contracts your AI product actually needs, in an area of Indian law that’s genuinely still being written, including by the courts, in real time.
Building your own model, fine-tuning an existing one, or building on top of a third-party API, share the details and we’ll come back with a clear scope and fixed fee.
Most startup legal categories are applying well-established law to a new business. AI is different, courts and regulators in India are actively working out how existing law applies to model training, AI-generated output, and automated decision-making, often in real time. That means an AI startup lawyer’s job isn’t just knowing the current rules, it’s tracking where they’re actually headed and structuring your contracts to survive the uncertainty. This page is part of our broader startup lawyer in India practice, focused specifically on what AI companies need.
Indian copyright law has no explicit exception for using copyrighted content to train an AI model, which has made this one of the most consequential open questions for any startup training or fine-tuning a model. On 24 July 2026, the Delhi High Court delivered India’s first substantive judicial ruling on the question, in ANI Media Pvt. Ltd. v. OpenAI Inc., a case where an Indian news agency alleged its copyrighted articles were used without authorisation to train the model behind ChatGPT.
The Court declined to grant an interim injunction, holding that the temporary storage of ANI’s content during training did engage the reproduction right under the Copyright Act, but was protected by the fair dealing exception at this stage. This matters, but it needs to be understood precisely: this was an interim finding. The Court explicitly stated its observations were provisional and would have no bearing on the final outcome, and the underlying suit is still ongoing. The government has separately convened an expert panel to assess whether the Copyright Act itself needs updating for AI. If your product trains on third-party content, this is exactly the kind of live legal question that needs a current, considered answer, not an assumption based on what another AI company is doing.
Indian copyright has traditionally required human authorship, which raises real questions for content, code, or designs your product generates with limited human input. For enterprise customers, this needs to be addressed explicitly in your terms: what they can do with AI-generated output, what you warrant about it, and who bears the risk if that output turns out to infringe someone else’s rights. Leaving this ambiguous is one of the more common gaps we find in AI product terms.
If your model is trained or fine-tuned on personal data, or your product makes automated decisions affecting real people, you’re within scope of India’s Digital Personal Data Protection Act. This gets stricter fast if your product could reach users under 18, the DPDP Act’s absolute restrictions on behavioural tracking and profiling of children apply regardless of your product’s general terms, see our EdTech startup lawyer page if that applies to you. Enterprise customers increasingly ask direct questions about training data provenance and data handling before they’ll sign, and having clear, honest answers ready is part of closing the deal, not just staying compliant.
If you’re building on an open-source model, its licence terms, usage restrictions, redistribution rules, attribution requirements, need to be reviewed properly, not assumed to be permissive because the model is free to download. If you’re building on a proprietary API, your commercial terms with that provider shape what you can promise your own customers. And if you’re licensing your own model or product out to others, your licence terms need to clearly define permitted use, liability limits, and what happens if your model is used in a way you didn’t intend.
Enterprise buyers of AI products increasingly negotiate specific terms most standard SaaS contracts don’t cover: IP indemnification if the AI’s output infringes a third party’s rights, clear liability caps for AI-specific risk, and disclosures about what data was used to train or fine-tune the model they’re buying access to. We draft these terms with your actual risk profile in mind, not a generic enterprise MSA with “AI” inserted into it. See our SaaS lawyer and MSA drafting pages for the underlying contract structures this typically builds on.
Terms covering what your product does, how outputs are generated, and clear limits on what you warrant about accuracy or fitness for purpose.
Licences or agreements covering data you use to train or fine-tune your model, structured to manage the legal risk this data actually carries.
Terms for licensing your model or product to others, or for your own use of a third-party model under its licence.
Standing terms for enterprise customers, including AI-specific liability and indemnification provisions.
Required wherever you process customer or end-user data as part of delivering your AI product.
Making sure models, code, and datasets built by founders, employees, and contractors actually belong to the company.
India doesn’t yet have a dedicated AI law comparable to the EU’s AI Act. What exists instead is a mix of general law, copyright, data protection, IT law, being actively tested against AI-specific facts, government advisories that continue to evolve, and active policy discussion about whether existing law needs updating. For a founder, that means compliance here isn’t a one-time checklist, it’s something that needs revisiting as the legal picture actually develops, which is exactly the kind of ongoing attention we build into how we work with AI clients.
AI investors scrutinise IP and data provenance closely during diligence, an unresolved training-data question can genuinely stall a round. We draft term sheets, SAFEs, and shareholders’ agreements for AI founders, and help you get diligence-ready on the IP and data side before investors start asking.
My Legal Pal is led by Prakhar Rai, an advocate enrolled with the Bar Council of India and an alumnus of NLSIU Bangalore, with a Master of Business Laws and 10+ years of experience, working with AI founders on the training data, IP, and contract questions this sector is still actively working out.
Our full startup legal practice, all stages.
For AI products built as a SaaS platform.
Check your platform’s data compliance gaps.
Tell us what you’re building, and we’ll come back with a clear scope and fixed fee.
