India to Charge OpenAI, Google for AI Training on Copyrighted Data

India Proposes Royalty System for AI Training on Copyrighted Data

A new proposal from India could significantly alter the operational landscape for AI companies like OpenAI and Google. The nation is considering a mandatory royalty system for firms that utilize copyrighted content to train their artificial intelligence models.

Framework for Access and Compensation

India’s Department for Promotion of Industry and Internal Trade has unveiled a proposed framework. This would grant AI companies access to all copyrighted works specifically for training purposes.

In return, these companies would be required to pay royalties to a newly established collecting body. This body will be comprised of organizations representing rights holders, and will then distribute payments to the original creators.

Addressing Concerns Over Copyrighted Material

The proposal centers around a “mandatory blanket license.” This is intended to reduce compliance costs for AI firms.

Simultaneously, it aims to ensure that writers, musicians, artists, and other rights holders receive compensation when their work is used to train commercial AI models.

Global Context and Legal Uncertainty

India’s initiative arises as global concerns mount regarding how AI companies train their models using copyrighted material.

This practice has already sparked legal challenges from authors, news organizations, artists, and other rights holders in both the U.S. and Europe.

The question of whether such training constitutes “fair use” remains unresolved by courts and regulators.

This legal ambiguity allows AI firms to expand rapidly without definitive regulations.

An Interventionist Approach

While the U.S. and the EU are focused on debates surrounding transparency and fair-use boundaries, India is proposing a more direct intervention.

The plan involves granting AI companies automatic access to copyrighted material, contingent upon mandatory royalty payments.

Committee Rationale and Benefits

An eight-member committee, formed by the Indian government in April, argues this system will circumvent years of legal disputes.

It will also guarantee creators are compensated from the beginning of the AI development process.

The committee’s 125-page submission details that a blanket license will “provide easy access to content for AI developers,” reduce transaction costs, and “ensure fair compensation for rightsholders.”

It is described as the most efficient method for managing large-scale AI training.

The submission further explains that the single collecting body will act as a “single window,” streamlining negotiations and facilitating royalty distribution to both registered and unregistered creators.

India’s Growing Importance in the AI Landscape

The committee highlights India’s increasing significance as a market for Generative AI tools.

Referencing OpenAI CEO Sam Altman’s statement that India is the company’s second-largest market, potentially surpassing the U.S., the committee argues that AI firms benefit substantially from Indian users.

Given their reliance on the work of Indian creators for model training, a portion of the generated revenue should be returned to those creators.

This rationale supports the establishment of a “balanced framework” that ensures adequate compensation.

Ongoing Legal Disputes

India’s proposal emerges amidst escalating legal battles globally concerning the lawful use of copyrighted material for AI model training.

In India, news agency ANI has filed a lawsuit against OpenAI in the Delhi High Court, alleging unauthorized use of its articles.

This case has prompted the court to examine whether AI training itself constitutes an act of reproduction or falls under “fair dealing” provisions.

Similar disputes are being addressed in U.S. and European courts, with creators asserting that tech companies have constructed their models using unlicensed content.

Concerns Arise Regarding India's AI Proposal

Despite the Indian government's recently unveiled proposal concerning artificial intelligence, significant reservations and opposition have been voiced.

Nasscom, the prominent industry association representing major technology companies such as Google and Microsoft, has formally registered its dissent. They advocate for India to implement a comprehensive text-and-data-mining (TDM) exception.

This exception would permit AI developers to utilize copyrighted material for training purposes, provided the source is legally obtained. Nasscom cautioned that a compulsory licensing system could hinder progress and suggested that copyright holders should have the option to exclude their content, rather than requiring payment for all data used in training.

The Business Software Alliance, representing global tech leaders including Adobe, Amazon Web Services, and Microsoft, has similarly urged the Indian government to steer clear of a licensing-exclusive approach.

They propose the introduction of a specific TDM exception, contending that depending solely on licensing for AI training data could prove unfeasible and suboptimal. Limiting AI models to only licensed or public domain data could negatively impact their performance.

BSA further warned that such restrictions could lead to reduced model quality and an increased risk of outputs mirroring the limitations and biases present in the restricted training datasets. A clear TDM exception, they argue, would strike a better balance between fostering innovation and protecting the rights of content creators.

The committee responsible for the proposal dismissed consideration of both a broad TDM exception alongside an opt-out mechanism. Their reasoning centered on concerns that such systems would either weaken copyright safeguards or prove unenforceable.

Instead, a “hybrid model” was put forward. This model would allow AI companies automatic access to all lawfully accessible copyrighted works, coupled with a requirement to remit royalties to a central collecting organization for distribution to creators.

Currently, the Indian government is soliciting public feedback on the proposal, providing a 30-day window for companies and stakeholders to submit their perspectives. The committee will then refine its recommendations based on this input before presenting the framework to the government for consideration.

Requests for comment from both OpenAI and Google have, as of yet, gone unanswered.