Representative image. | Photo credit: Getty Images/iStockphoto
The Union government is working to implement changes to the copyright law that will require new artificial intelligence requirements over the next three years, a senior official said on Thursday (Dec 11, 2025).
The Department of Industrial Promotion and Internal Trade (DPIIT) released a working paper last week on AI and copyright issues. He proposed a “free licensing” framework where websites whose data goes through large language models (LLMs), such as ChatGPT, would receive royalties through an author association that would split the money between them.
The proposal aims to address the growing tension worldwide between publishers of online content (such as book publishers and news organizations) and AI firms that “train” their LLMs with large amounts of textual data, typically downloaded from the public Internet. Publishers argued that they should be compensated because they contribute to the development and improvement of these models.
DPIIT’s proposal generally proposes that AI developers download content while ensuring that they eventually pay content publishers.
Payout after launch
Another working paper will follow in about two months, looking at whether AI-generated works are copyrightable and how their authorship is decided, DPIIT Additional Secretary Himani Pande said at a press briefing. Following this, the government is likely to introduce an amendment to the Copyright Act 1957 in Parliament to introduce the new regime.
Ms Pande said the copyright company, known as the Copyright Royalties Collective for AI Training (CRCAT), would demand payment from AI firms once they commercialize their models, as opposed to when they mine data from the internet to develop or train them.
Compensating copyright holders has been a contentious issue, with news publishers in particular suing AI companies around the world. Newswire agency ANI (from which The Hindu syndicates video content) and The New York Times are among those who have sued ChatGPT developer OpenAI for allegedly refuting their content in chat conversations. OpenAI denied their allegations.
AI firms disagree
The technical industry body Nasscom, which had one seat on the committee that drafted the report, disagreed with the model proposed by DPIIT. The body, which represents Google, Meta, Amazon and other big firms with significant AI investments, said publishers should be able to opt out of including their data in training models, warning that a blanket licensing model could expose firms to further litigation.
In a note shared with The Hindu, a major technology firm that develops AI models expressed specific concerns about the DPIIT proposal. “Under the cited copyright jurisprudence, the burden of proof to prove infringement is on the claiming copyright owner,” the note states. “The hybrid proposal reverses this. It is well established in copyright law that in a copyright infringement action, the plaintiff (content owner) must prove that the defendant (AI developer) (infringed their copyright). If the burden of proof is on the AI developer, they must prove that they did not use the content owner’s material, even if the output is similar. This is extremely difficult and technically infeasible, as Genistic AI tools are likely.”
Ms. Pande said that the input of companies will be taken into account during the consultation process.
Published – 11 Dec 2025 21:08 IST
