Using public data to train ChatGPT not commercial activity: OpenAI to HC

Sam Altman-led OpenAI on Tuesday told the Delhi High Court that its use of publicly available data to train its large language model (LLM) for ChatGPT is not a commercial activity in itself.

 

The court was hearing the plea moved by ANI Media against ChatGPT maker OpenAI in which the former has alleged that its content was illegally used to train LLMs and other artificial intelligence (AI) models by OpenAI.

 

ChatGPT argued that training the LLM is both a commercial and non-commercial activity.

 

"My private use of my publicly available works for research, including training my LLM, is not in and of itself a commercial activity. It will be commercial if I provide that as a service to others," said OpenAI's counsel.

 

 

It also argued that since ChatGPT drives traffic to the ANI website, there is no commercial harm being caused to the website.

 

The AI firm also argued that copyright law protects the creation and expression of content, not the discovery of ideas and facts.

 

Justice Amit Bansal recorded the submissions and listed the matter for May 16 to hear submissions from other parties.

 

In an earlier hearing, ANI told the court that ChatGPT used its material to train the AI software, which is then stored and general access is provided to ANI’s content.

 

During the proceedings, amicus curiae Adarsh Ramanujan said that language models were not designed as ‘truth machines’ but operated based on predictive algorithms. He said the language model did not use ANI’s data directly and ANI’s content was often published behind a subscription wall, yet its subscribers, who had varying levels of paywall restrictions, could republish the content.

 

Amicus curiae (literal translation ‘friend of the court’) is a person or organisation that offers information or advice to a court, even though they are not a party to the case.

 

Addressing the legal implications of public accessibility, advocate Ramanujan argued that the mere availability of content in the public domain did not nullify copyright protections.