Delhi High Court Rules AI Training on News Content Is Fair Dealing

DNI SUMMARY — KEY POINTS

• The Delhi High Court rejected a request for an interim injunction by Asian News International against the artificial intelligence firm OpenAI.
• Justice Amit Bansal ruled that the process of using copyrighted news material for training large language models constitutes fair dealing under Indian law.
• The court concluded that OpenAI does not infringe on intellectual property rights as the AI models do not reproduce content in a substantially similar manner.
• Legal experts view this as a landmark decision that provides temporary protection for the growth of generative artificial intelligence development across India.
• Ongoing litigation will continue to address deeper questions regarding copyright ownership and the long-term impact of AI on the traditional media industry.

The Delhi High Court has delivered a pivotal interim ruling in the copyright dispute between Asian News International and OpenAI, establishing that the use of published news material to train large language models falls under the fair dealing exception. Justice Amit Bansal emphasized that restraining the development of artificial intelligence at this stage would be detrimental to public interest and the broader technological ecosystem. By classifying the data ingestion process as a form of non-expressive use, the court has provided a significant boost to companies currently developing generative tools within the Indian market.

Fair Dealing Under Indian Law

The legal framework surrounding artificial intelligence remains a subject of intense debate as traditional intellectual property protections clash with rapid technological innovation. In its decision, the court noted that the current Copyright Act must be interpreted in a way that balances the rights of creators with the transformative benefits of machine learning research. This ruling suggests that courts are increasingly viewing the ingestion of massive datasets as a foundational necessity for artificial intelligence rather than a straightforward violation of existing literary copyrights held by major media organizations.

Concerns raised by ANI centered on the unauthorized scraping of its original reporting and the potential for ChatGPT to regurgitate verbatim content, thereby threatening its business model and editorial credibility. The news agency argued that such large-scale ingestion constitutes commercial exploitation of intellectual property without proper licensing or remuneration. While these arguments were presented with vigor during thirty-two separate court sittings, the judge remained unconvinced that an immediate injunction was necessary to prevent irreparable harm while the case proceeds toward a final resolution.

Balancing Innovation And Intellectual Property

The implications of this judicial stance extend far beyond this specific legal conflict, influencing how global technology firms approach data sourcing in South Asia. By confirming that the storage of data for training purposes does not constitute an automatic infringement of the Copyright Act, the court has lowered the threshold for AI developers navigating the complexities of Indian law. This decision mirrors similar, though often contradictory, international debates regarding the intersection of copyright law and the evolution of sophisticated, data-hungry generative platforms.

Retrieval-Augmented Generation processes were also a critical focus for the court, as plaintiffs argued these systems could effectively replace human-curated news websites by providing real-time summaries. However, the court found no evidence that the outputs generated by OpenAI were substantially similar enough to the original news reports to warrant a finding of infringement. This distinction between the process of internal model training and the generation of final responses remains a vital component of the defense mounted by the artificial intelligence industry globally.

Technology And The Media Industry

Critics of the decision suggest that this ruling could leave news publishers in a precarious position, struggling to monetize their content in an era dominated by automated information retrieval systems. The potential for these systems to hallucinate or misattribute facts remains a primary worry for legacy media houses that rely on accuracy and reputation to maintain their subscription and advertising revenue. Even as the legal battle continues, the dialogue between software companies and traditional publishers is likely to remain tense as both sides look for sustainable commercial models.

Legislative bodies are already reacting to the uncertainty exposed by this case, with the Commerce Ministry considering the formation of an expert panel to evaluate the robustness of current copyright regulations. The government recognizes that the existing statutes may not explicitly account for the unique characteristics of generative artificial intelligence and the massive data requirements that fuel them. This move underscores the urgency of creating a clearer policy roadmap that supports both technological advancement and the financial stability of the news industry.

Future Of Copyright And Regulation

Future proceedings will likely shift toward more technical scrutiny of how large language models are trained and whether specific guardrails can prevent the inadvertent reproduction of protected material. As Justice Bansal noted, the societal benefits derived from scientific and technological research remain a relevant factor for judicial consideration in modern contexts. While this interim ruling provides a sense of clarity for now, the ongoing legal saga will continue to shape the contours of digital intellectual property rights for years to come.