A landmark interim ruling by the Delhi High Court could provide significant legal cover for AI developers to train models on copyrighted content under the 'fair dealing' exception. Expert Arul George Scaria breaks down the implications.
- The Delhi High Court held that training LLMs on copyrighted works may fall under India's 'fair dealing' exception.
- The ruling stems from the ANI vs. OpenAI copyright infringement lawsuit.
- This provides a legal shield for developers like OpenAI and Sarvam AI.
- Liability may still exist regarding the actual output generated by the AI.
In a significant development for the global technology landscape, the Delhi High Court has issued an interim ruling that could reshape how Artificial Intelligence (AI) is developed in India. The court suggested that using copyrighted works to train Large Language Models (LLMs) could potentially fall within the scope of India's 'fair dealing' exception.
The legal debate was ignited by a lawsuit filed by news agency Asian News International (ANI) against OpenAI. ANI alleged that the creators of ChatGPT infringed upon their copyright by using their proprietary content to train their AI models without authorization. While the matter is subject to appeal, the interim decision offers a measure of legal protection to major players like OpenAI and indigenous developers such as Sarvam AI.
Why This Matters
BozokMedia analysis shows that this ruling is a critical pivot point for India's digital economy. Had the court ruled strictly against AI training, it might have stifled the growth of domestic generative AI innovation, leaving India trailing behind global competitors.
"This is a balanced and forward-looking judgment that encourages innovation, though it does not grant a blanket immunity for the output side of LLMs." - Arul George Scaria
Arul George Scaria, a law professor at the National Law School of India University (NLSIU) and an expert in intellectual property, served as an amicus curiae in the case. He highlighted that the Delhi High Court took a dynamic approach to interpreting 'private use' and 'research' within the fair dealing framework.
Crucially, the court distinguished the Indian legal framework from the US 'fair use' doctrine. While US courts utilize a broad four-factor test, the Indian 'fair dealing' provision is narrower and requires a two-stage analysis: first determining if the use falls under specific permitted purposes, and second, conducting a fairness analysis.
Legal Comparison: India vs. USA
| Feature | India (Fair Dealing) | USA (Fair Use) |
|---|---|---|
| Scope | Narrow and Specific | Broad and Open-ended |
| Testing Method | Two-stage (Purpose + Fairness) | Four-factor test |
| Flexibility | Statutory-based exceptions | Highly flexible/Judicial-led |
The court specifically addressed the 'market harm' aspect, questioning whether LLM responses would act as a direct substitute for ANI's syndication services. The court concluded that they would not, thereby upholding the public interest in technological advancement.
Frequently Asked Questions
1. Does this mean AI companies don't have to pay for data?
Not necessarily. While training might be covered under fair dealing, the legal landscape is evolving and final decisions may differ.
2. Can an AI be sued for copyright infringement?
Yes, if the AI's output reproduces copyrighted material in a substantial way, the developer could face liability.
