ANI versus OpenAI: Is ‘open-source’, large language model AI above intellectual property norms?
India’s first lawsuit against the use of copyrighted material to train an AI chatbot has all the makings of a landmark case, writes Harsh Gour.
Harsh Gour
Published on: 9 December 2024, 03:18 pm
ON November 19, 2024, the Delhi High Court heard India's first lawsuit against OpenAI, the company behind the famous artificial intelligence (AI) chatbot ChatGPT.
The court issued a summons to OpenAI following a lawsuit filed by Asian News International (ANI). ANI accuses OpenAI of using its copyrighted news content without permission to train its large language model (LLM), ChatGPT. ANI is seeking ₹2 crore in damages and an injunction to stop OpenAI from using its content in future.
According to ANI, OpenAI uses the information on its web platform without permission. ANI contends that OpenAI has no authority to use its content for profit, even though it is freely accessible online. This issue is significant because it draws attention to the continuous tension between the intellectual property rights of content providers and data utilisation by AI businesses.
Background to the case
ANI is one of the major news agencies in India. It provides content for many media channels across the country. OpenAI is accused of having trained its language model, ChatGPT, on news content belonging to ANI without prior authorisation and while being fully aware that it has no authority to do so.
ANI also contends that OpenAI ignored the agency’s request for a licensing agreement. ANI seeks fair remuneration and believes that OpenAI should comply with India’s copyright laws.
The defendant, OpenAI Incorporation, is a San Francisco-based firm that has created ChatGPT and other AI systems. The company is well-known in the tech industry for training its AI models on large datasets. OpenAI contends that its technology creates human-like responses by analysing patterns and deriving implications from vast databases rather than just replicating specific pieces of material.
OpenAI argues that because they were using data publicly available on the internet, they should not require permission to use such data.
Claims and counterpoints: The debate unfolds
ANI's position
During the hearing on November 19, ANI's counsel, advocate Sidhant Kumar, argued that OpenAI had violated India’s copyright laws by training AI models on ANI content. ANI’s content can only be accessed or stored by a limited number of people even though it is available online, he contended.
In addition, Kumar argued, web scraping for training AI models without consideration for compensation constitutes a serious theft from the companies that produce such content, and it is within their right to say how their content should be used.
He contended that ANI had offered OpenAI licensing on October 3, 2024, but the US-based company had declined the offer. ANI also pointed out that OpenAI's refusal of licensure shows its utter disrespect for intellectual property protection principles and copyright laws.
