Publishing contracts that allow AI chatbots to learn from books

29 July 2023

A few weeks ago, I wrote about two authors, Mona Awad and Paul Tremblay, who had filed a law suit against OpenAI, the developer of ChatGPT. Awad and Tremblay were claiming books they had written were being used to help “train” the AI powered chatbot. They say this had happened without their prior knowledge or permission.

It now looks like there may be a solution to this problem, but not perhaps the one writers have been seeking. According to a tweet by the Australian Society of Authors (ASA), some book publishers in the United States are adding clauses to their publishing contracts, allowing the works of authors to be used to train generative chatbots:

We know that some terms of service in publishing have already included clauses allowing the use of authors’ work to train AI and we are now hearing that authors in the US are being asked by publishers to agree to clauses which allow their work to be used to train generative AI.

That’s sure as hell one way to “solve” the problem. But I wonder if authors agreeing to their works being used in this fashion are being offered additional remuneration? And what of writers who disagree with such a proviso? Do their works go unpublished?