Delhi High Court rules OpenAI's use of ANI content to train ChatGPT wasn't copyright infringement as the news agency didn't show ChatGPT reproduced its reports
An Indian court said on Friday that OpenAI's use of news agency ANI's content to train its ChatGPT service did not amount to copyright infringement.
Context & Ripple Effects
ANI's case arrives amid a widening set of publisher challenges to AI training: The New York Times' landmark suit against OpenAI and Microsoft put alleged training on millions of articles at the center of the dispute, while an earlier US case involving Raw Story and AlterNet was dismissed for insufficiently shown harm. The Delhi ruling matters because it turns on a concrete boundary in this record—whether ChatGPT reproduced ANI reporting—rather than treating model training alone as dispositive.
First-order effects
- OpenAI avoids copyright liability in ANI's claim on the court's finding that ANI did not establish reproduction of its reports by ChatGPT.
- ANI loses this route to restrict or obtain relief for OpenAI's use of its reporting, absent a successful challenge or a different factual showing.
Second-order effects
- News publishers bringing training claims may face greater pressure to document allegedly infringing model outputs and resulting harm, not merely the presence of their work in training material.
- The ruling gives OpenAI a favorable decision to cite in licensing and litigation discussions, even as its disputes elsewhere remain fact- and jurisdiction-specific, including the US fight over production of anonymized ChatGPT logs.
Third-order effects
- If courts continue to distinguish training from demonstrable output reproduction, copyright disputes over generative AI may increasingly focus on output testing, provenance evidence, and remedies for retrieval or memorization.
- Divergent rulings across jurisdictions could preserve uncertainty over whether publishers can rely on litigation to shape AI-content licensing, reinforcing commercial negotiations alongside court cases.
The trend: Generative-AI copyright litigation is moving from broad objections to training toward evidence-intensive tests of whether models reproduce protected content and cause cognizable harm.