Indian Court Says OpenAI’s Use of News Content for AI Training Is Not Copyright Infringement

Lead

According to OSChina, the Delhi High Court ruled on July 24 that OpenAI’s use of content from Asian News International (ANI) to train artificial intelligence models does not constitute copyright infringement. Justice Amit Bansal stated that the conduct falls within a research-related fair use exception under India’s Copyright Act. The court also found that ANI had not submitted sufficient evidence showing that ChatGPT responses reproduced ANI’s protected content.

Key Points

• The dispute centered on training data. ANI’s claim focused on whether OpenAI’s use of news material during model training violated copyright. The court’s analysis turned on whether that use could be protected by an exception under copyright law.
• Research-based fair use played a central role. The ruling indicates that AI model training may, in certain circumstances, be examined under a research exception. This does not mean every unlicensed use of copyrighted material is automatically lawful, but it does show that purpose and legal context matter.
• Evidence of output copying remains crucial. The court noted that ANI had not adequately shown that ChatGPT outputs copied protected expression from its content. In copyright litigation involving generative AI, proving that material was used for training is different from proving that the model produced infringing output.

Why It Matters

The ruling is significant not only because OpenAI prevailed in this particular case, but also because it addresses a broader question now facing courts around the world: can copyrighted or news content be used to train generative AI systems? In this instance, the Indian court appeared willing to recognize a legal space for training under a research-oriented fair use framework.

For AI developers, the decision may provide some reassurance, but it is not a universal shield. Copyright exceptions vary widely across jurisdictions, and companies deploying models globally still face uncertainty around data sourcing, permissions, and compliance.

For publishers and news agencies, the case highlights the evidentiary challenge in AI copyright claims. It may not be enough to argue that content entered a training corpus; plaintiffs may also need to show a concrete connection between protected expression and allegedly infringing outputs. As a result, negotiations over licensing, transparency, and compensation between media organizations and AI companies are likely to continue.

Source: OSChina