The Ministry of Electronics and Information Technology (MeitY) has backed a statutory licensing framework for training artificial intelligence (AI) systems on copyrighted works, its submission to the Department for Promotion of Industry and Internal Trade’s (DPIIT) committee on AI and copyright reveals.
Notably, it has rejected both unrestricted data access and consent-based licensing models, warning that AI’s public benefits cannot operate as legal immunity from copyright obligations.
In its submission to the DPIIT committee, MeitY endorsed its proposed hybrid AI licensing model. The framework allows AI developers to train on copyrighted works without prior permission, while requiring them to compensate rights holders through a centralised, revenue-linked royalty mechanism.
Notably, the submission places the government at odds with technology industry groups that have argued for lawful access-based training without licensing. Instead, MeitY frames AI training as a process of value aggregation built on human creative labour, rather than as a purely technical activity insulated from copyright law.
MeitY’s submission leads to a central policy question: How to support large-scale AI development without undermining the economic and moral foundations of creativity, or treating AI as a legal exception?
Public benefit does not override copyright claims: MeitY
MeitY opens its submission by presenting AI as a potentially transformative public good. It points to AI’s role in improving healthcare outcomes, agricultural productivity, access to education, and public service delivery.
Moreover, the Ministry acknowledges that many scientific and healthcare breakthroughs enabled by AI depend on aggregating vast amounts of human knowledge and effort. It argues that policymakers have a clear imperative to remove barriers where such benefits can be widely deployed.
However, MeitY draws a firm legal boundary, remarking that public benefit alone cannot determine legality.
“Public good, while a worthy objective, cannot become a qualifying criterion for innovation, or unconditional immunity against legal challenges,” MeitY’s submission states.
In effect, MeitY accepts the case for enabling AI at scale. At the same time, it rejects the idea that societal benefit can automatically override copyright claims: signalling resistance to treating AI as a special case under copyright law.
How MeitY reframes copyright in the AI era
Unlike industry submissions that characterise AI training as a purely technical process, MeitY places copyright at the centre of the AI value chain. The submission treats copyright as recognition of human creativity and labour, arguing that generative AI has disrupted traditional ideas of authorship by shifting value from the creative process to the output. Crucially, MeitY contends that this disruption strengthens, rather than weakening, the case for copyright protection.
Furthermore, by describing creative works as a shared digital topography shaped by collective effort, MeitY advances a collectivist understanding of creative contribution. This framing supports compulsory access to training data paired with statutory compensation, instead of individual, consent-based permissions.
This approach underpins the Union Ministry’s resistance to unrestricted training access sans compensation, while also legitimising a shift away from creator-level control over training permissions.
Why does MeitY still prioritise broad data access?
MeitY acknowledges that building an indigenous and self-reliant AI ecosystem requires access to large, diverse, and high-quality datasets.
The Ministry directly links data access to model accuracy, bias mitigation, and representational fairness. According to the submission, inclusive AI systems depend on broad participation from the populations they aim to serve.
However, MeitY warns that excessive compliance burdens could undermine these objectives. It argues that such burdens could limit participation and slow innovation. This concern explains the Union Ministry’s rejection of consent-based and opt-out licensing models: which it implicitly treats as unworkable at scale, particularly given enforcement and transparency constraints.
The central issue, as MeitY frames it, is not whether AI systems should train on copyrighted works, but how the value generated through such training should be distributed.
Advertisements
Why MeitY backs statutory licensing and revenue-based compensation
To address this tension, MeitY has endorsed the DPIIT committee’s hybrid model of statutory licensing combined with revenue-linked compensation.
Under this framework, AI developers may train on copyrighted works without seeking individual permissions. However, once their systems cross a prescribed revenue threshold, they must share revenues with rights holders.
Additionally, the Ministry has recommended that the proposed Copyright Regulatory and Compensation Authority for Training (CRCAT) define a minimum revenue floor for royalty obligations. It has also urged CRCAT to adopt a proactive approach for allocation and dispute resolution, positioning the authority as an active governance body rather than a passive royalty-clearing mechanism.
And instead of treating copyright as a right to exclude, this model reconfigures it as a right to remuneration, administered through a centralised regulatory authority with significant discretion over implementation.
How this differs from industry submissions
MeitY’s position diverges sharply from that of industry groups such as the Business Software Alliance (BSA). These groups argue that developers should be free to train AI systems on any lawfully accessed content, with copyright law stepping in only when AI-generated outputs infringe protected works.
Industry submissions frame training as the extraction of non-copyrightable patterns and relationships. Based on this view, they contend that licensing requirements misunderstand how AI systems operate and risk slowing innovation.
MeitY, by contrast, accepts output-stage remedies but insists that the law must also recognise value extraction at the training stage once AI systems generate commercial returns, even when no single act of infringement can be clearly identified.
This divergence exposes a deeper divide over whether copyright law should intervene only at the point of harm in AI outputs, or also regulate how creative value is accumulated during training.
What questions remain unresolved
While MeitY endorses the DPIIT committee’s hybrid approach, its submission raises several unresolved questions.
Most notably:
- Who will set the value of creative works used in AI training, and how will authorities measure that value across different kinds of creators and content?
- Can statutory remuneration genuinely replace consent once creators lose control over how their works are used?
- How much authority will the CRCAT hold over pricing, distribution, and disputes? And what checks will limit that power?
Overall, these questions highlight the central trade-off in MeitY’s position. The submission prioritises scale and predictability through centralised governance, but leaves unclear how India will balance power, value, and accountability in its AI ecosystem.
Also Read:
Support our journalism by subscribing
For YouSource link
