The New York Times is ‘not telling the full story’
OpenAI responds to lawsuit, challenging accusations of copyright infringement | Newspaper ‘cherry-picked’ its regurgitation evidence | Complaint met with ‘surprise and disappointment’ by the AI company.
OpenAI has responded to a The New York Times (NYT) complaint with a strongly worded statement claiming the publication is “not telling the full story”.
NYT accused OpenAI and Microsoft of copyright infringement on December 27, 2023, alleging the pair had copied “millions” of its articles to train their chatbots.
The newspaper further alleged that OpenAI and Microsoft used this work to develop and commercialise their generative AI products without obtaining its permission.
However, in a blog post published yesterday, January 8, OpenAI said the lawsuit came as a “surprise and disappointment”, claiming it first heard about the action by reading it in the NYT.
Discussions with NYT had been “progressing constructively” in their last communication on December 19, 2023, said the AI company.
The blog post explained that the negotiations held focused on a “high-value partnership” around real-time display with attribution in ChatGPT.
The proposed partnership aimed to offer NYT a novel way to connect with both their existing and new readers, while providing ChatGPT users access to the publication's reporting.
OpenAI considered the lawsuit to be “without merit” and offered arguments to counter those put forward by NYT.
OpenAI fights back
According to the blog, OpenAI explained to NYT that their content did not significantly “contribute to the training of existing models and wouldn’t be impactful for future training.”
The AI research organisation expressed disappointment that NYT failed to share specific examples of content regurgitation despite their “commitment to investigating” and addressing any concerns.
OpenAI emphasised its dedication to addressing issues promptly, as evidenced by the removal of a ChatGPT feature in July that could inadvertently reproduce real-time content.
OpenAI contended that the regurgitations attributed to it by NYT seemed to be years-old articles that have proliferated on third-party websites.
It alleged that the prompts used by NYT may have “intentionally manipulated” the model to regurgitate content—such as including lengthy excerpts of articles to get the model to regurgitate.
OpenAI claimed that its models don’t typically behave in the way insinuated by NYT, suggesting that the examples provided were “cherry-picked” or the model was instructed to regurgitate.
Despite NYT's claims, OpenAI asserted that this misuse is not typical or allowed user activity and should not be considered a substitute for the publication.
However, the ChatGPT developer hoped for a “constructive partnership” with the newspaper in the future.
Defending its commitments
OpenAI additionally sought to defend itself by explaining its collaboration with news organisations to create new opportunities.
OpenAI's objectives included using its products to assist reporters and editors by streamlining time-consuming tasks, educating its models about the world, and presenting real-time content with attribution in ChatGPT, as a way to allow news publishers to engage with readers.
It argued that training AI models on publicly available internet materials is fair use, however, offer an opt-out process for publishers to prevent its tools from accessing their sites—the NYT adopted this in August 2023.
Finally, OpenAI argued that regurgitation and memorisation is a “rare failure” of the learning process but the company is working to “drive to zero.”
Did you enjoy reading this story? Sign up to our free daily newsletters and get stories sent like this straight to your inbox
