With the growing number of lawsuits against AI companies over the unauthorised use of content in the training of AI models, one of the leading firms, OpenAI, is now offering artists a way to prevent their works from being used in AI model training. However, this solution seems late.

The Tussle Between Creators and AI Firms

AI programs, whether it’s for image generating or chatbots, are trained using terabytes of content scraped from the internet. However, some of the companies behind these models have done so using content obtained without permission, basically stealing the works of human creators and artists. 

The situation has since resulted in multiple lawsuits and articles against the likes of OpenAI for potentially using copyrighted works to illegally train AI models. 

In September, over a dozen authors filed a lawsuit against OpenAI for infringing on their copyrights by using their books to train its popular ChatGPT chatbot. “The success and profitability of OpenAI are predicated on mass copyright infringement without a word of permission from or a nickel of compensation to copyright owners,” the complaint reads. 

A month prior, Cryptopolitan also reported that lawyers for the NY Times were also considering suing the ChatGPT creator due to copyright infringement.

OpenAI’s Tacit Response to Copyright Issues

As fears and legal pressure about the encroachment of AI on creative industries intensify, OpenAI quietly rolled out the latest version of its image-generating AI program, DALL-E 3, with the option to allow creators to prevent their works from being used in training AI models. 

Also, DALL-E 3 would now decline requests that ask for images designed in the style of an artist. “DALL-E 3 is designed to decline requests that ask for an image in the style of a living artist. Creators can now also opt their images out from training our future image generation models,” OpenAI said.

While this is appreciably a step in the right direction for artists to protect their works going forward, however, it may also not make much difference because the contents published before 2023 could have already been fed into the AI dataset, and it is not economically feasible for companies to retrain their programs due to individual opt-out requests.

“The past? It’s done—most of it, anyway,” said Daniel Gervais, a law professor at Vanderbilt University who specialises in AI and copyright.