# Delhi HC rules AI training can be fair dealing in ANI case

2026-08-06T21:51:59+00:00 | Governance | Indian Opinion Desk

Corroboration: 1 independent outlet

The Delhi High Court has refused to block OpenAI from using copyrighted news content to train its ChatGPT model, marking India’s first major judicial ruling on AI and copyright. In the case brought by Asian News International, Justice Amit Bansal held that training an AI model may qualify as 'private or personal use, including research' under Section 52(1)(a) of the Copyright Act. The court rejected ANI’s request for an injunction because the similarities between ChatGPT outputs and ANI’s articles were 'inadequate' to prove infringement at this stage. Legal experts are divided on what the order means. Some say it effectively creates a text-and-data-mining exception that Parliament never wrote, while others stress the ruling is fact-specific and interim. The court also applied a three-factor test: whether the data was used only for training, whether it harms the news business, and whether the AI serves public interest. ANI is expected to appeal, and the Hindustan Times notes that the final trial will decide whether large language models actually 'memorise and regurgitate' substantial parts of protected works.

## Coverage

- legal.economictimes.indiatimes.com <https://legal.economictimes.indiatimes.com/news/industry/delhi-hc-ruling-on-ai-training-a-transformative-shift-in-copyright-law/132850784>
- hindustantimes.com <https://www.hindustantimes.com/opinion/no-news-content-for-ai-training-without-licences-101786032949239.html>

Tags: AI copyright, ANI, Delhi High Court, fair dealing, OpenAI
Canonical: https://indianopinion.org/delhi-high-court-ruling-puts-ai-training-under-fair-dealing-test/
License: Summary and commentary (c) Indian Opinion, reusable with attribution. Facts belong to the linked sources.
Cite: https://indianopinion.org/delhi-high-court-ruling-puts-ai-training-under-fair-dealing-test/#story-in-brief

Review state: restored archive article, accurate at time of publication, not offered to search indexes.
