DPIIT’s framework for AI copyright faces questions on transparency, consent, and compliance costs

Paul McCartney, titan of modern songwriting, released a track that contained nothing but recordings of empty studios in November. An unusual protest, McCartney joined thousands of musicians who spoke against the UK government’s planned changes to copyright law, which would make it easier to train AI models on copyrighted work without a licence. 

“The album, titled Is This What We Want?, featured recordings of empty studios and performance spaces, representing the impact on artists’ and music professionals’ livelihoods that is expected if the government does not change course,” the statement from McCartney said. 

Not just the UK, AI copyright is a globally significant issue that has gained critical attention as the technology continues to impact every facet of life and work. Primarily, AI copyright concerns itself with two points: the complex legal question of who owns creative works made with AI, and whether work generated by AI itself qualifies for copyright.

The Department for Promotion of Industry and Internal Trade (DPIIT) released a working paper on AI and copyright issues last week. One of the major takeaways from the paper is a ‘blanket licensing’ framework for the use of ‘lawfully accessed copyright-protected works’ to train AI systems, in exchange for remuneration to the creator. The catch, however, is a clause that accompanies this framework. “Under this framework, the rights holders will not have the option to withhold their works from use in the training of AI Systems,” the paper notes.

The Ministry of Electronics and Information Technology (MeitY) has endorsed the proposed framework, noting that the ‘model has the potential to equitably meet its multifaceted objectives across the domains of technological innovation and creative labour’.

The non-opt-out nature of said blanket licensing has caused concern about pursuants of creative work and the industry in general. They argue that it could undermine creators and artists’ choice and agency. Notably, IT industry body Nasscom expressed its dissent against this approach, saying, “rights holders should be provided clear statutory protection.” The body creators should have simple and clear ways to opt out – either through a tag for public content or through licence terms for non-public content.

“Weakening author’s rights”

Experts opine that this framework, which can be construed as forcing creators with no choice, can open doors for possible legal scrutiny, especially from publishers, artists or collecting bodies who already license their work commercially.

“Copyright law is built on consent and control, not just compensation. A blanket licence that allows use by default, even with payment, could be challenged on the grounds that it weakens an author’s exclusive rights. In most copyright regimes globally, including India’s, the right to authorise or refuse use is central,” said Devroop Dhar, Co-Founder & India CEO, consulting firm Primus Partners.

There is also concern about how ‘lawfully accessed’ can be interpreted in this context. Public visibility alone does not mean lawful access. Content behind paywalls, subscription models or usage-restricted platforms would still require authorised access, and scraping at scale could violate contractual terms even if individual access is legal.

“Indian publishers, especially regional and vernacular ones, operate on thin margins. If lawful access is interpreted loosely, many will respond by restricting access or moving content behind paywalls. That would be counterproductive for India’s open internet and multilingual knowledge base. Precision in definition is essential to avoid unintended effects,” said  Rishi Agrawal CEO and Co-Founder of Teamlease Regtech.

This brings to another core challenge about how and whether there is a procedure for original creators to know if their work has been used for AI training, especially those without institutional backing. AI training pipelines are typically opaque, and models lack a clear, auditable record of the individual works used in training.

“A credible framework would typically require recordkeeping + disclosure + auditability,” said Jagdish Bhandarkar, Partner at Deloitte India. He suggests including developer submissions detailing training data sources and categories, the maintenance of robust logs, and the establishment of independent audit or inspection mechanisms. According to the legal analysis, the Working Paper envisions creating a centralised authority (CRCAT) and a disclosure mechanism, along with a database where creators can register their works for royalty allocation.

“The goal should be balance. A system with clear transparency, flexible pricing, and real opt-out choices can protect creators without slowing innovation. If done well, it could make India an attractive place to build responsible AI. If done poorly, it risks pushing creators and smaller players to the margins,” said Jaspreet Bindra, co-founder of consultancy AI&Beyond.

Risks to smaller startups

As of mid-2025, Nasscom reports India's Generative AI (GenAI) startup ecosystem has seen explosive growth, reaching over 890 startups, a 3.7X surge from the previous year, positioning India as the world's second-largest hub for GenAI apps. 

Without careful design, the impact could be asymmetric. Large global firms can absorb royalties and compliance overhead. Smaller Indian startups cannot, say experts.
A blanket licence may seem like it reduces legal risk on paper, but it may be seen as shifting risks. If royalty rates, data declarations, or dispute handling are unclear, it just adds a new compliance layer.

“For smaller Indian AI startups, this can hurt more through extra fees, paperwork, and reporting which can slow them down and strain budgets. Big global firms will absorb it with large legal/compliance teams. For them, this is just another cost of doing business,” said Pawan Prabhat, Co-Founder of genAI firm Shorthills AI.

A system with clear transparency, flexible pricing, and real opt-out choices can protect creators without slowing innovation for AI developers and organisations. If these concerns are taken into account, it could make India an attractive place to build responsible AI, the absence of which risks pushing creators and smaller players to the margins.