MeitY backs DPIIT framework on copyright use for AI training

The proposed model seeks to balance large-scale AI development with creator rights by allowing licensed data use for training, introducing revenue-based compensation, and recommending safeguards to protect startups while ensuring transparency in royalty distribution.

The Ministry of Electronics and Information Technology (MeitY) has backed a framework proposed by a DPIIT-led committee to govern the use of copyrighted material in training artificial intelligence systems. The endorsement comes amid growing global legal disputes, where authors and publishers have challenged major technology firms over the unauthorised use of protected content for building large language models.

In its observations attached to the government’s working paper on AI and copyright, MeitY said the proposed structure attempts to strike a practical balance between encouraging innovation and recognising the economic value of creative work. The ministry noted that unrestricted access to data is critical for developing accurate and inclusive AI systems, but such access must be accompanied by fair compensation for rights holders.

Hybrid licensing model for AI training

At the core of the proposal is a hybrid approach that combines mandatory blanket licensing for AI training with a statutory right to remuneration. Under this model, AI developers would be allowed broad access to copyrighted datasets, while creators would be entitled to payments linked to the commercial success of AI systems.

MeitY said the framework has the potential to address the competing needs of rapid technological advancement and protection of creative labour. By tying compensation to revenues, the model aims to ensure that creators benefit when AI products move from experimentation to commercial deployment.

Revenue thresholds and transparency in royalties

To prevent early-stage companies from being overburdened, the ministry has recommended that a minimum revenue threshold be set. Only once AI developers cross this threshold would revenue-sharing obligations apply. This, MeitY said, would encourage innovation and experimentation while still capturing value once AI solutions mature.

The ministry also stressed the importance of a transparent and robust mechanism for collecting and distributing royalties. It warned that unclear rules could lead to disputes and litigation, urging the proposed Copyright Royalties Collective for AI Training to adopt a clear, participatory approach in defining how payments are calculated and shared.

While MeitY has endorsed the DPIIT framework, industry body Nasscom has expressed reservations, favouring a text and data mining exception with opt-out provisions. The differing views highlight the complexity of shaping AI policy that supports both innovation and intellectual property rights.