The U.S. Copyright is seeking feedback from the public surrounding AI and copyright issues. Here’s the latest.
The Copyright Office announced that it wants the public to answer three main questions about AI. How copyrighted data should be treated when training AI models, whether AI-generated material should be copyrightable, and how copyright liability would work when AI assisted in the development of a piece. The office is also specifically seeking comments around AI potentially violating publicity rights when used to mimic voices, likenesses, or artistic styles.
Ready to share your opinion with the copyright office? You can submit written comments until October 18, while any replies must be submitted by November 15. Whether or not AI-generated works can be copyrighted is a subject of hot debate. The office says applications for copyrighted works with AI-generated materials have been increasing in recent years, so it wants public feedback on how to handle the rise of AI.
Meanwhile, several artists have taken to suing OpenAI and Meta for using their copyrighted works to train large language models. Comedian and author Sarah Silverman is one such creator.
Her lawsuit alleges that her books were used to train these LLMs from “illegally acquired datasets containing the works available in bulk via torrent systems.” ChatGPT is able to provide a summarized version of Silverman’s book, among many others. The lawsuit notes the chat bot does not reproduce any of the copyright management information that Plaintiffs included with their published works.”
Both lawsuits against OpenAI and Meta claim that plaintiffs did not authorize the use of their copyrighted works as training material for any language models built by either company. Each lawsuit contains six counts of copyright violations, negligence, unjust enrichment, and unfair competition. Authors are looking for statutory damages, restitution of profits, and more. But it also raises an important question—does an illegally trained LLM have a place in society?