India AI copyright

India's AI-Copyright Policy Is Drifting to the Courts While DPIIT's Part 2 Paper Sits Unwritten

DPIIT's AI-copyright committee reportedly has not met since June, so ANI v OpenAI, not policy, may set India's rules on AI training and outputs.

India's Stalled AI-Copyright Framework People of Internet Research · India 8 Dec 2025 Part 1 released Part 2 has not followed. 8 Committee members Set up by DPIIT on 28 April 2025. 8 Dec 2026 Next ANI appeal hearing Delhi HC division bench; some outl… peopleofinternet.com
India's Stalled AI-Copyright Framework People of Internet Research · India 8 Dec 2025 Part 1 released 8 Committee members 8 Dec 2026 Next ANI appeal hearing peopleofinternet.com

Key Takeaways

India's most consequential AI-copyright proposal has stalled halfway. According to a Hindustan Times report relayed by DigWatch, the Department for Promotion of Industry and Internal Trade (DPIIT) expert committee on AI and copyright has not met since June 2026, and Part 2 of its working paper has still not been prepared. DPIIT did not comment. We could not open the DigWatch page, and we did not locate the Hindustan Times original, so the meeting claim rests on that relayed report. What we can verify independently is how long the gap has run: Part 1 came out on 8 December 2025, and as of today, 21 November 2026, Part 2 has not.

What Part 1 proposed

DPIIT set up the committee on 28 April 2025, according to Bar and Bench's report of a Lok Sabha reply, which describes an eight-member body. Part 1, per Storyboard18's coverage, recommends a hybrid model: a mandatory blanket licence for AI developers, with royalties at government-set rates that are open to judicial review, and centralised collection and distribution. Creators would receive equal compensation whether or not they belong to a collecting society. The paper is hosted on the DPIIT site, and DPIIT announced it through PIB.

The strongest case for this design deserves a fair statement. Indian creators and publishers have little bargaining power against well-funded model developers. A statutory licence removes the transaction costs of negotiating with millions of rights-holders. It also guarantees payment without letting any single holder block training. That is a coherent answer to a real market failure, and the committee was right to reject a zero-price exception that would leave writers and musicians uncompensated.

Why the unfinished half matters more than it looks

Storyboard18 reports that Nasscom lodged a dissent recommending a full text-and-data-mining allowance with opt-out mechanisms, while MeitY supported the hybrid model. So the committee itself was split on the central question, and the government never resolved that disagreement publicly. A split paper followed by a stalled committee leaves developers with a proposal they cannot plan around and no decision to either adopt or abandon it.

Part 2 was meant to settle the other half of the problem. Search coverage of the paper's scope indicates it would address whether AI outputs can be copyrighted, who the author is, whether moral rights apply, and who is liable for infringement. Those questions decide whether a startup building on a foundation model can own what its product generates and who is on the hook when an output copies a protected work. They are at least as commercially important as the training question.

The delay is not new in kind. In December 2025, Deccan Chronicle quoted a senior government official saying the committee expected to release its second paper in about two months. Bar and Bench's Lok Sabha report from the same month described Part 2 as "still under review." Neither timeline has held. Nearly a year after Part 1, nothing has replaced it.

The courts are filling the vacuum

When a legislature or ministry does not decide, litigation does. In ANI Media Pvt. Ltd. v. OpenAI, LiveLaw reports that a Delhi High Court division bench of Justices Avneesh Jhingan and Manmeet Pritam Singh Arora issued notice on ANI's appeal against the denial of interim relief and listed it for 8 December. Other outlets have reported 5 December, so the cause list should be checked. Per LiveLaw, the single judge held that OpenAI's storage of ANI's works did not amount to infringement and treated training as covered by the research defence. Bar and Bench confirms the court sought OpenAI's response to the appeal.

That matters for two reasons. First, an interim-relief ruling on the Copyright Act, 1957 will be read as a signal on a question Parliament has not addressed, and the Act's fair-dealing provisions were not drafted with model training in mind. Second, a court deciding one news agency's dispute cannot design a royalty regime, set rates, or settle output authorship. It can only say yes or no on infringement in front of it. Courts produce binary outcomes; the policy problem needs a graduated one.

A proportionate way forward

The pro-innovation reading is not that Part 1 should be dropped. It is that India should decide. Three steps would help:

A ministry that publishes half a framework and then goes quiet imposes costs without providing rules. Developers cannot know whether training on Indian content will carry a statutory charge, and rights-holders cannot know whether they will be paid. If the 8 December hearing produces a substantive signal first, India will have let a single appeal bench set national AI-copyright policy by default. That is a poor way to govern a technology the government itself wants to promote.

Sources & Citations

  1. LiveLaw: Centre proposes mandatory blanket licence for AI training (DPIIT Working Paper Part 1)
  2. European Commission IP Helpdesk: India's 'One Nation, One Licence, One Payment' AI-copyright framework
  3. Bar and Bench: Copyright Act being reviewed, Centre in Lok Sabha
  4. Storyboard18: DPIIT working paper proposes hybrid licensing model
  5. Deccan Chronicle: DPIIT panel to release second working paper in 2 months
  6. LiveLaw: Delhi HC issues notice on ANI's appeal against OpenAI
  7. Bar and Bench: Delhi HC seeks OpenAI reply on ANI plea