Data for training stored overseas, copyright law doesn’t apply: OpenAI

OpenAI asserts that its language model is engineered to prevent verbatim reproduction, claiming that training on news agency content constitutes a non-extractive, non-expressive use permissible under Indian copyright law. The company argues that because the data processing occurs on servers outside India, local copyright statutes do not apply, and that extracting factual elements for model development does not infringe upon the rights of news organizations. The case, initiated by ANI, highlights the tension between AI development and journalistic copyright protection. OpenAI contends that facts themselves are not subject to exclusive rights, while ANI maintains that unauthorized scraping and training for commercial gain violate existing copyright protections. This dispute has drawn support from various media and publishing groups, emphasizing the broader industry concern over the commercial exploitation of original journalistic work without licensing. This litigation is critical to open_data ecosystems as it sets a potential precedent for how AI companies can legally access, process, and utilize publicly available or licensed datasets. The outcome will likely define the boundaries of fair use in machine learning, influencing whether developers must obtain explicit permissions for data scraping or if transformative, non-expressive uses remain lawful. Such rulings directly impact the scalability and legal security of open data initiatives within the AI sector.

Source: hindustantimes.com
Published on 2025-03-30

Data for training stored overseas, copyright law doesn’t apply: OpenAI

Related news