Delhi High Court Ruling: Redefining 'Fair Use' Boundaries in AI Training Data Copyright Disputes

I noticed an interesting detail: when the Delhi High Court ruled that OpenAI's use of ANI content for AI training did not constitute infringement, it invoked the concept of "non-expressive use." This term isn't common in NLP, but in the context of copyright law, it perfectly captures the statistical nature of text processing during AI training—the model doesn't "read" or "understand" articles; it extracts the underlying distributional features.