Sarah Silverman’s copyright suit against OpenAI, the company behind ChatGPT, might not have a favorable outcome for the comedian. Silverman alleges that OpenAI violated her rights by using her book, The Bedwetter, to train ChatGPT, which can summarize parts of the book. However, a closer look reveals that there might not be a clear violation in this case.
Although OpenAI’s ChatGPT was able to provide a detailed summary of Malcolm Gladwell’s book, The Tipping Point, its summary of chapter 1 of The Bedwetter is vague. This could suggest that OpenAI has instructed ChatGPT to avoid summarizing the book to avoid potential copyright issues.
To examine the legality of OpenAI’s actions, we need to consider whether a copyright violation occurs when a computer reads a book. According to legal expert Deven Desai, OpenAI’s behavior with books is akin to Google’s digitizing of millions of books, which the Supreme Court deemed legal. Paraphrasing and summarizing content are not copyright violations, as evident from the presence of book summaries on platforms like Amazon.
While it is unfortunate for copyright owners, such as Sarah Silverman, the reality is that OpenAI’s actions may not breach copyright laws. Settling the lawsuit would invite countless other legal challenges from copyright owners worldwide. This situation raises questions about fair compensation for copyright owners.
One possible solution could be to impose a small fee that large language models like ChatGPT would pay to copyright owners for the use of their content. However, even if this fee amounted to billions of dollars, individual compensation for copyright owners would be minimal, similar to the challenges faced by musical artists with streaming.
Alternatively, legislative bodies like Congress could address this issue through new copyright laws. However, major revisions to copyright laws pass infrequently, with the last significant revision being the Digital Millennium Copyright Act in 1998. It is unlikely that AI-related copyright revisions will occur within the next five years, especially considering the lack of a legal framework for AI tools that are trained on publicly available content.
Unfortunately, authors like Sarah Silverman may find themselves in a challenging position, as there seems to be limited recourse available to address their concerns regarding AI’s use of copyrighted material. Ultimately, this situation underscores the need for further conversations and potential revisions to copyright laws in the future.
