India has no single AI law yet. Companies building or using AI are governed by existing laws: the Digital Personal Data Protection Act, 2023, the Copyright Act, 1957, the Information Technology Act, 2000 and the rules under it, and ordinary contract law. Some important questions, especially on copyright, are still unsettled.
This checklist covers the legal issues that most often come up for Indian companies building AI products, or using third-party AI tools inside the business.
Personal data in training and prompts
If you train, fine-tune or run a model on personal data of people in India, the DPDP Act applies to that processing. You need a lawful ground, usually consent or a legitimate use under section 7, and the notice, security, retention and rights duties that come with it. Prompts and uploads from users that contain personal data are also processing.
Section 3(c)(ii) says the Act does not apply to personal data that is made or caused to be made publicly available by the data principal to whom it relates, or by someone under a legal obligation to publish it. Read this narrowly. It does not cover all data found on the internet. Data that someone else posted about a person, or data that was leaked, is not covered by the exception. You need to be able to show where the data came from.
Section 17(2)(b) exempts processing for research, archiving or statistical purposes, but only if the data is not used to take decisions specific to a data principal and the processing follows the standards in the Rules. Most commercial model development will not fit easily within it. Section 8(3) also requires reasonable efforts to ensure data is complete, accurate and consistent where it is used to make a decision that affects the person.
Copyright in training data and outputs
Whether training a model on copyright works without permission infringes copyright in India is unsettled. The fair dealing provisions in section 52 of the Copyright Act are narrow and purpose-specific, and no Indian court has yet given a final answer on AI training. Litigation brought by Indian publishers against an AI developer is pending before the Delhi High Court. Until courts or Parliament decide, treat the question as open.
Ownership of AI outputs is also uncertain. Section 2(d)(vi) treats the author of a computer-generated work as the person who causes it to be created, but whether a given output is original enough to be protected, and who 'caused' it, will depend on the facts. Practical steps help regardless of how the law develops:
- Prefer licensed, public-domain or your own data for training, and keep records of sources.
- Respect website terms and technical opt-outs when collecting data.
- Filter outputs for close reproduction of training material.
- Say clearly in your terms who owns outputs, and that you give no guarantee they are protectable.
- Seek IP indemnities from model providers where you can.
IT Act duties and MeitY advisories
If your AI product hosts or transmits user content, such as a chatbot inside a social platform or an image tool where users share outputs, you may be an intermediary under the IT Act and must follow the IT Rules, 2021, including content rules, a grievance officer and takedown duties. Generating deepfakes or impersonating people can also attract offences under the IT Act, including section 66D (cheating by personation using a computer resource) and section 66E (violation of privacy).
The Ministry of Electronics and Information Technology (MeitY) has issued advisories to intermediaries and platforms on deepfakes, AI models and labelling of synthetically generated content, and has moved to address synthetic content through the IT Rules. These have been revised over time. Check the current position before launch, and build labelling, user warnings and complaint handling into your product so you can adapt.
Your terms of use and disclaimers
- Explain that outputs are generated by AI and may be wrong, incomplete or offensive.
- Set acceptable-use rules: no unlawful content, impersonation, or harmful use.
- Say whether you use prompts and outputs to train models, and how users can opt out.
- Allocate ownership of inputs and outputs.
- Limit liability fairly; disclaimers do not override consumer protection law for consumers.
- For high-stakes uses such as health, finance or legal information, add clear warnings and human review.
Contracts with AI vendors
| Issue | What to look for |
|---|---|
| Data use | Whether the vendor can use your prompts, files and outputs to train its models; ask for a no-training commitment for business data. |
| Confidentiality | Treatment of your inputs as confidential information, retention periods and deletion on request. |
| Personal data | Processor terms under the DPDP Act, sub-processors, location of processing and breach notice. |
| Security | Certifications, access controls and incident reporting fast enough for your CERT-In and DPDP duties. |
| IP indemnity | Whether the vendor defends you against claims that the model or its outputs infringe third-party rights, and the conditions attached. |
| Output ownership | A clear statement that you own, or may freely use, the outputs. |
| Liability | Caps, carve-outs, and what happens if the vendor changes or withdraws the model. |
An internal AI-use policy
Employees already use AI tools. A short written policy reduces the risk of confidential information, source code or customer personal data being pasted into tools that keep it. Cover which tools are approved, what data may never be entered, when outputs need human review, how to label AI-generated material shared with clients, and who to ask. Train staff on it and review it every six months.
How we help
Prodigy Legal reviews AI product terms, data sourcing and DPDP compliance, negotiates AI vendor contracts, and drafts internal AI-use policies. The Tech & IP monthly plan covers SaaS and licensing contracts, DPDP and IT Act compliance, and IP questions as they come up.
Frequently asked questions
Is there an AI law in India?
Not a dedicated one, as of the date of this guide. AI is governed by existing laws, including the DPDP Act, the IT Act and IT Rules, the Copyright Act and contract law, along with MeitY advisories. Check for new rules before launch.
Can I train an AI model on publicly available data in India?
Only partly settled. The DPDP Act does not apply to personal data that the person themselves made public, but that exception is narrow. Copyright in scraped content is a separate, unsettled question. Keep records of sources and take advice on your specific dataset.
Who owns AI-generated content in India?
The Copyright Act treats the person who causes a computer-generated work to be created as its author, but whether a particular output is protected is uncertain. Settle ownership between the parties by contract.
Do I need to label AI-generated content?
MeitY has issued advisories, and has moved to address synthetic content through the IT Rules. Requirements depend on your role and the current notified text, so check before launch. Labelling is good practice in any case.
Last reviewed: 2026-11-22. This guide gives general information about Indian law for companies. It is not legal advice, and every matter turns on its own facts.