Read Time: 11 minutes
Senior Advocate Amit Sibal, in the last hearing, argued that, “Facts which are news of the day cannot be protected under copyright laws. The first person who posts the fact has not created the fact but has discovered it first. The discoverer finds and records but does not create that fact”.
ChatGPT, on Tuesday, submitted before the Delhi High Court that copyright law protects the creation and expression of content, not the discovery of ideas and facts. These arguments were made in a plea filed by ANI Media, alleging copyright infringement by ChatGPT, an AI-based language model.
The bench of Justice Amit Bansal observed that the matter would be decided based on the broader framework of the statute rather than its granular details. The court recorded the submissions and listed the matter for May 16, 2025.
Senior Advocate Kapil Sibal virtually appeared in the matter, representing the Broadband India Forum as an intervenor. Broadband India Forum is an independent think tank that promotes policy and regulatory initiatives to enhance India’s broadband ecosystem, working closely with the government and stakeholders.
Senior Advocate Amit Sibal commenced his arguments by asserting that there existed no copyright over quotations. He submitted that Indian copyright law distinguished ideas and facts available in the public domain, and the law permitted the use of expressions derived from such sources. He argued that copyright protects creation, not discovery, and that mere adaptation or rearrangement of facts does not amount to infringement unless it involves expression protected by law.
Senior Advocate Amit Sibal emphasized that the outputs generated by ChatGPT were not rearrangements of ANI’s expressions, but entirely new expressions based on statistical modeling. He explained that ChatGPT’s large language model (LLM) was trained up to a particular cut-off date. Post that, it accessed publicly available data through Microsoft’s search engine. He clarified that ChatGPT provided sources and attribution when offering information beyond the training data’s scope.
“It is an LLM and not a search engine, and so when it gives information post the cut off date, it also provides a link to the source along with attribution”, Senior Advocate Amit Sibal added.
Senior Advocate Amit Sibal further distinguished between probative similarity and substantial similarity, submitting that ANI had failed to demonstrate the latter, which was essential to establish actionable copyright infringement. He stressed that even if access and probative similarity were established, substantive similarity must still be proven.
“Probative similarity is used to determine if copying has occurred, while substantial similarity is used to determine if the copying is substantial enough to be considered actionable infringement”, Senior Advocate Amit Sibal submitted.
On the issue of data storage, Senior Advocate Amit Sibal acknowledged that copies were made during the training of the LLM. However, he argued that only non-expressive elements such as grammar, syntax, facts, and knowledge structure were extracted—elements not protected by copyright. He illustrated this by stating that sentences like “the moon is the Earth’s only natural satellite” were broken down into subject, object, and verb structures, which the machine learned to process logically.
Senior Advocate Amit Sibal further contended that copyright law was never intended to prevent the extraction of common linguistic patterns or factual information. The act of copying, he argued, occurred merely as an intermediary step to teach the LLM and was not meant for human consumption. He maintained that this step did not amount to infringement under Section 14 of the Copyright Act.
Senior Advocate Amit Sibal also argued that a literal interpretation of Section 14(1)(a), which mentions “storage”, would be inappropriate. He advocated for a purposive interpretation of the law in light of technological advancements, stating that earlier courts had upheld similar principles. Drawing a comparison, he submitted that just as copying by hand was not impermissible before the invention of photocopiers, technological replication for learning should not be treated differently.
“Before the invention of the photocopier, students would have to copy by hand- if copying by hand was permissible, how could using a photocopier be impermissible?”, Senior Advocate Amit Sibal argued.
Senior Advocate Amit Sibal also addressed the issue of tokenization and vectorization, explaining that these were technical processes used by machines to break down language for internal understanding. These processes, he stated, did not constitute adaptations or infringing reproductions of content.
When questioned by the court whether the use of ANI’s content could fall under fair use provisions, Senior Advocate Amit Sibal responded affirmatively, adding that a fair-dealing analysis would support ChatGPT’s position. He also noted that ANI failed to demonstrate any actual market impact caused by ChatGPT’s functionality.
On the question of liability for incorrect responses, Senior Advocate Amit Sibal pointed to multiple disclaimers issued by ChatGPT. He clarified that the AI did not position itself as a news syndicate or authoritative source and required users to independently verify the information.
Senior Advocate Amit Sibal argued that although ChatGPT’s services were accessible in India, the model itself had no physical presence, office, or business operation within the country, unlike other entities such as Amazon, which conceded to conducting business in India.
For ANI: Advocates Sidhant Kumar, Akshit Mago, Manyaa Chandok, Om Batra, Anshika Saxena and Shagun ChopraFor OpenAI: Senior Advocate Amit Sibal with Advocates Sanjeev Kapoor, Nirupam Lodha, Madhav Khosla, Gautam Wadhwa, Moha Paranjpe, Vanshika Thapliyal, Ankit Handa, Darpan Sachdeva, Saksham Dhingra, Rajat Bector, For Intervenors: Advocates Ameet Datta, Ayush Hoonka, Riddima, Naimish Tewari, Akshay Natrajan and Harsh KaushikCase Title: ANI Media Pvt Ltd v Open AI
Please Login or Register
