Battle over whether information public or private

Many discussions around artificial intelligence (AI) are currently taking place in India. These include the need to build homegrown large language models (LLM), the drafting of voluntary standards for the fair development of AI, the setting up of an AI safety institute, and the costs and benefits of providing AI computing services under the IndiaAI programme.

Even as policy debates rage on one front, an AI battle has opened up on another. This is in the form of a copyright case brought in the Delhi High Court by ANI Media, an Indian news agency, against OpenAI. The local company has alleged that OpenAI stored and used its data, mainly its news articles, to train LLMs without permission. It also accused ChatGPT, OpenAI’s generative AI chatbot, of not only reproducing its content verbatim but also attributing false statements to ANI. OpenAI has defended these claims on grounds of jurisdiction, maintaining that it neither operates nor stores data in India. It also denied that information was presented to the user in the same form as the source. Asserting that copyright law protects expression, not facts or ideas, OpenAI denied it had infringed ANI’s copyright.

Because the case raises a number of new legal issues not previously dealt with by Indian courts, two amici curiae, or friends of the court, have been appointed. They will help to answer a number of questions the court will have to address. These include whether the use of the data involved for training purposes amounts to copyright infringement and whether fair use is made out. The latter permits limited use of copyrighted works without permission for criticism, comment, news reporting, teaching, scholarship and research. The last issue is whether Indian courts have the necessary jurisdiction. Expected to take time to resolve, the case will certainly lay down novel AI law.

OpenAI has blocked ANI’s domain and excluded it as a source of training data going forward, in keeping with its opt-out policy. This is a Band-Aid for a bullet wound. The policy requires ANI to indicate whether it wishes to opt out of OpenAI using its data to train AI tools. If every authorised user or crawler were required to exercise its opt-out option, it would be logistically impractical.

This case raises important basic questions regarding AI LLM training data. As well as copyright issues, such use of proprietary data raises concerns of wrongful attribution, dilution of reputation and antitrust.

The court has the unenviable task of balancing the protection of intellectual property rights against the frenetic pace and inevitability of innovation and progress. It is critical to ask how far courts can go. The court could confine its findings to the facts of the case, or it could attempt more sweeping declarations about AI chatbots and training data.

The court may, however, prefer to urge the legislature and executive to develop a clear regulatory framework, one that allows innovation to flourish while protecting creativity and originality in today’s digital world.

The questions before Delhi High Court may be new for India, but are not unique globally. Other jurisdictions are facing similar issues. For example, in the United States, OpenAI faces similar allegations from The New York Times newspaper (NYT), which claims that its proprietary content was included in the training data used to develop ChatGPT. Besides undermining its ability to monetise content, NYT has argued that this is an unlicensed and unauthorised use, amounting to copyright infringement. It has also accused OpenAI of attempting to profit from its substantial investment in journalism without fair compensation. Similar conflicts are being fought in Canada and in the United Kingdom, where major news outlets and individual authors have accused OpenAI of using their content without authorisation or compensation.

Defending its approach as transformative, by adding new expression or meaning to the original material, OpenAI has stated it intends to work collaboratively with news organisations while advancing public access to information. Whether these developments are viewed as simple disputed copyright infringements or as complex, multi-sided debates that will define the future of AI, innovation and human progress, there is no doubt that these are indeed exciting IP times.

Essenese Obhan is the managing partner and Nandini Chowdhry is an associate at Obhan & Associates.