Can AI use Copyrighted Works?

As Artificial Intelligence (AI) becomes a daily tool, a massive legal question is emerging: can AI companies freely read and learn from news articles, books, and images protected by copyright, or is that stealing?

The Delhi High Court recently issued a landmark interim ruling that addresses this exact clash between technology and creator rights. In a high-profile case between Asian News International (ANI) and OpenAI (the creator of ChatGPT), the Court declined to stop OpenAI from using ANI's news content to train its AI models.

What Were the Facts of the Case?

Asian News International (ANI), a major Indian news agency, filed a lawsuit against OpenAI. ANI alleged that OpenAI had used its copyrighted news reports without permission to "train" ChatGPT. According to ANI, OpenAI unlawfully stored its copyrighted works and generated responses that either copied ANI's original reporting or falsely attributed incorrect information to the news agency.

To protect its business while the main lawsuit was ongoing, ANI asked the Delhi High Court for an "interim injunction"—a temporary emergency order to completely stop OpenAI from using its content while the trial took place.

The core legal question before the Delhi High Court was: Does storing and using a publisher's copyrighted news articles to train an Artificial Intelligence model amount to copyright infringement, or is it legally permitted as "fair dealing"?

Why Did the Delhi High Court Refuse to Stop ChatGPT?

At this early, interim stage of the lawsuit, Justice Amit Bansal declined to grant the emergency injunction against OpenAI. The Court found no immediate proof of copyright theft and made several key observations:

1. Training AI is "Fair Dealing"

Under Indian copyright law, using someone's work for specific purposes like research or education can be exempt from infringement under a rule called "fair dealing" (specifically Section 52(1)(a) of the Copyright Act, 1957). The Court held that, at first glance, storing and using ANI's works strictly to train large language models falls within this fair dealing exception.

2. No Proof of "Memorizing" the News

ANI failed to prove that ChatGPT had actually memorized and "regurgitated" exact copies of its copyrighted articles. The law protects the specific expression of an article, not the underlying facts or news events themselves.

3. AI Responses Were Not Copies

The Court examined responses generated by ChatGPT using a technique called Retrieval-Augmented Generation (RAG). The judge found that the AI's generated answers were not substantially similar to ANI's original literary works.

4. Protecting the Public Interest

The Court highlighted that granting a blanket injunction at this early stage would not only harm OpenAI but also severely affect the broader public interest in accessing and developing technological innovations like AI.

Why Is This Judgment Important?

This judgment is the first of its kind in India and will heavily influence future legal battles between AI developers, media organizations, authors, and tech companies. It enters a massive, ongoing international debate where courts worldwide are struggling to balance the rights of traditional publishers with the need for technological progress. By interpreting AI training as potentially falling under "research" and "fair dealing," the Court has temporarily provided a much-needed breathing space for AI development in India.

The Key Takeaway

This judgment does not declare that every use of copyrighted material for AI training is automatically legal, nor does it create a blanket immunity for tech companies. Instead, it offers an important initial interpretation of the Copyright Act, recognizing that training an AI is different from simply copying a book. The final, definitive legal answers will only be decided after a full trial, but for now, AI training continues.

Frequently Asked Questions (FAQs)

Is it legal for AI to read copyrighted content? According to this interim ruling, storing and using copyrighted material strictly for the purpose of training an AI model may fall under the "fair dealing" exception for research, meaning it is not automatically considered copyright infringement.

  1. Can I sue ChatGPT if it writes an article that looks like mine? If ChatGPT generates an output that is a direct, substantial copy of your uniquely expressed work, it could be infringement. However, copyright protects how you write, not the raw facts or ideas you write about. You must prove the AI actually copied your specific expression, not just the information.
  2. Did ANI lose the case completely? No. This was only an interim (temporary) ruling on whether to immediately stop OpenAI while the trial continues. The main lawsuit will still proceed to a full trial, where both sides will present detailed evidence.

If you are a publisher, author, or tech developer dealing with copyright disputes, AI licensing, intellectual property rights, or dealing with unauthorized commercial use of your content, obtaining case-specific legal guidance is crucial to protecting your creative assets in the digital age.

NyayNiti provides comprehensive legal assistance in intellectual property law, copyright infringement, and technology-related litigation. Contact NyayNiti to discuss your legal concerns and understand the appropriate, strategic options available to safeguard your rights.

Disclaimer: This article is intended for general legal information and public awareness and does not constitute legal advice for a specific case. The applicable law and available remedies depend on the unique facts and procedural history of your situation. This article discusses the interim Delhi High Court judgment in ANI Media Pvt. Ltd. v. OpenAI OpCo LLC, decided in July 2026.