By The realm of generative artificial intelligence (AI) is not without its complexities, particularly when it comes to the question of whether comprehensive large-language models can be developed without infringing upon the rights of content creators.
A striking example cited in the OpenAI lawsuit involves a request made to ChatGPT to summarize Silverman’s book “The Bedwetter.” Remarkably, the AI model provides a detailed summary of the book—an outcome the lawsuit argues would not have been possible without the incorporation of the complete text into the training model.
Lawsuits Filed Against Use of Unlicensed Copyrighted Material in Training Sets
By Eric J. Savitz
While Meta declined to comment on the lawsuits, OpenAI has yet to respond to a request for comment.
Most language model creators provide limited information about the underlying data supporting their models. However, Adobe stands out as an exception, explicitly stating that its image-creation model exclusively uses licensed images from its stock-photography business or images no longer under copyright protection. Additionally, Adobe is actively working on a method for creators to opt out of having their works included in training models and to receive compensation if the software generates new work based on their creations.
Comments