India's government wants to set prices for the content AI companies use to train models

The government of India wants AI companies to pay for accessing content they use to train models, but only once they start producing revenue.

That idea emerged yesterday in a working paper [PDF] prepared by a Committee on Generative Artificial Intelligence and Copyright formed by India's Department of Promotion of Industry and Internal Trade.

The paper notes that developers of AI models mostly didn't pay for copyrighted content, and the global debate about fair use exemptions to copyright law that followed.

The Department concludes that free access to content - a "zero price license model" - is not appropriate because it "would undermine incentives for human creativity and could lead to long-term underproduction of human generated content."

The Committee's members also found "access to large volumes of data and high-quality data is crucial for AI development" but fears negotiations to license that content could lead to "long negotiations and high transaction costs [which] can hold back innovation, particularly for startups and MSMEs."

It therefore proposed a hybrid model that has the following three elements:

The paper even suggests a name for the royalty collection organization - The Copyright Royalties Collective for AI Training (CRCAT) - and recommends it be a nonprofit organized by associations of rightsholders. It also proposes the establishment of a "Works Database for AI training royalties" that would invite content creators to register their works in order to be eligible to receive royalties from CRCAT.

"By preserving the right of the copyright owners to receive royalties and administering it through a single umbrella organization made by the rightsholders and designated by the government, the model aims to provide an easy access to content for AI Developers for AI Training, simplify licensing procedures, reduce transaction costs, ensure fair compensation for rightsholders," the paper states.

Precedents exist for the proposed arrangement. Several countries operate performing rights organizations that collect royalties from venues such as restaurants and retailers that play recorded music. Those royalties are pooled and disbursed to artists. Your correspondent is a member of an Australian scheme that charges royalties for reprints of news and other content and disburses them to creators who register their works.*

India's government has declared the nation will become a world leader in all aspects of AI, an ambition that sees it take a mostly friendly attitude towards tech giants as they address the local market. Tech giants, however, continue to argue fiercely for the right to train their models without first paying for content - but are also doing deals that cover their ongoing operations.

India, however, poses a considerable challenge because the nation recognizes 22 scheduled languages, eight of which are spoken by over 50 million people, and has a huge and fragmented media and publishing ecosystem. This proposal may therefore go down well with Big Tech, if New Delhi makes royalty payments worth their while. ®

*I end up with a couple of hundred dollars a year but other journos I know - mostly those who work in finance media - have told me they can score thousands a year.