SAN FRANCISCO – OpenAI, the company behind the popular AI chatbot ChatGPT, has responded publicly to a recent copyright infringement lawsuit filed by The New York Times. In a blog post, OpenAI said the lawsuit is “without merit” and that the Times “is not telling the full story.”
The lawsuit, filed in late December, alleges that ChatGPT reproduced Times articles verbatim when given certain prompts. OpenAI disputes this claim, saying the Times likely manipulated the prompts intentionally to generate copied text or cherry-picked isolated examples after many tries.
“Even when using such prompts, our models don’t typically behave the way The New York Times insinuates, which suggests they either instructed the model to regurgitate or cherry-picked their examples from many attempts,” the blog post states.
OpenAI says it has made efforts to reduce verbatim repetition in its AI systems like ChatGPT. The company claims the Times refused to provide specific examples of copied content before moving forward with legal action. The verbatim text included in the lawsuit seems to be from old articles already circulating widely online rather than directly from the Times, according to OpenAI.
The post does acknowledge that OpenAI removed a ChatGPT feature called Browse last year after finding it accidentally reproduced third-party content in some cases. But the company maintains training AI on vast amounts of text data, including copyrighted online content under fair use allowances, is essential for the technology’s development.
“In order for AI systems to solve new problems, represent the diversity of human knowledge and experience, and avoid harm, they need to be trained on the aggregate of human knowledge – which necessarily includes the enormous corpus of copyrighted works,” OpenAI’s post explains.
OpenAI recently made this argument about fair use and the need for broad AI training data access to UK lawmakers as well. Though the company says it respects copyright law and has offered publishers ways to opt out, it believes scraping websites without permission falls under fair use protections.
Last August, facing pressure from writers and publishers, OpenAI allowed websites to block its web crawling bots used to gather AI training data.
Despite the simmering legal battle, OpenAI says it hopes to one day partner with the Times, much like its existing content partnerships with Axel Springer and The Associated Press.
“We are hopeful for a constructive partnership with The New York Times and respect its long history,” the blog post concludes.
With ChatGPT’s meteoric rise in popularity in recent months, lawsuits and debates over AI copyright issues seem poised to continue. OpenAI’s response makes clear it believes broad access to data is crucial for further advancement in this fast-moving field of technology – even copyrighted data lacking explicit permission.
