The Delhi High Court has rejected a preliminary injunction request by Indian news agency Asian News International (ANI) against OpenAI, ruling that ANI failed to prove copyright infringement. The court found that the articles ANI submitted as evidence were published after the models had been trained, and thus could not have been part of the training data. Judge Amit Bansal denied the requested relief, stating that the similarities between ANI's content and ChatGPT outputs likely stemmed from Retrieval Augmented Generation (RAG), which allows models to retrieve information in real time. The ruling underscores the court's focus on the legal status of AI training and the public benefit of language models for education, research, and accessibility. Source: thedecoder
ANI submitted several ChatGPT outputs to the court, claiming they were substantial copies of its articles. However, OpenAI demonstrated that the models used, GPT-4 and GPT-4o, were trained on data from April 2022 and April 2024, while the articles ANI cited were published in August and September 2024. The judge noted that ANI's evidence did not support its claim that OpenAI permanently stores training data and can reproduce its work verbatim on demand. The court will revisit this issue in the main proceedings. Source: thedecoder
The court also found that ANI failed to demonstrate that AI training constituted copyright infringement. Both parties agreed that OpenAI used ANI content during training, but the judge emphasized that the material made up a tiny share of the overall dataset and that the model extracted only non-expressive elements such as grammar and syntax. The judge relied on an exception under Indian copyright law covering 'private or personal use, including research,' which he interpreted broadly to include AI training. The ruling marks the first time a court has explicitly found that AI training falls under a private use exception. Source: thedecoder
