OpenAI Accuses New York Times of ‘Manipulating’ ChatGPT

    Published on:

    OpenAI hits back at The New York Times, claiming the news organization “deliberately manipulated” ChatGPT and “regurgitated” entire lines from the newspaper's article, the developer defended the practice in a copyright lawsuit did.

    On December 27, The Times filed a lawsuit against OpenAI and its major investor Microsoft, alleging intellectual property infringement related to the use of millions of “proprietary” articles to train ChatGPT. filed a lawsuit.

    According to a filing in the U.S. District Court for the Southern District of New York, the newspaper sued OpenAI and Microsoft for “unlawful copying and use of the Times' unique and valuable copyrighted works.” They are seeking compensation for damages.

    Also read: OpenAI offers $1 million to news publishers to train LLMs with content

    Newspapers “don't tell the whole story”

    OpenAI said the lawsuit lacks “merit.” blog post The article, published by the AI ​​company this week, The New York Times added, “doesn't tell the whole story.” The developer claimed he first learned about the lawsuit in a news article published by The Times a few days after Christmas.

    “We work with news organizations to create new opportunities. Training is fair use, but we offer an opt-out because it's the right thing to do,” OpenAI wrote. The Times adopted the content removal option in August, but continued to pursue the lawsuit months later, the newspaper said.

    In that copyright case, the news organization claimed that ChatGPT “regurgitated” many of its articles. This is the tendency for AI chatbots to spit out entire “memorized” passages of a particular section of content or articles. The newspaper is calling on OpenAI to destroy training data and AI models that use copyrighted material without consent.

    OpenAI explained in a blog post that regurgitations are “a rare bug and we are working to eliminate them.” But the company also accused the paper of choosing a convenient prompt that was intentionally designed to provoke regurgitation rather than normal customer usage. The examples cited by the Times in the lawsuit are from old articles published on multiple third-party sites, the newspaper said.

    “guess so [the Times] “To regurgitate our model, prompts are intentionally manipulated and often include long excerpts from articles,” the company said.

    “Our models typically don't behave in the way the New York Times alludes to. This suggests that we either told them to regurgitate them, or that we picked a sample from many attempts. Masu.”

    Ian Crosbie, a partner at law firm Susman Godfrey, which represents the paper, said: Said The Financial Times said, “In this blog, we acknowledged that OpenAI used The Times' work and many other works to build ChatGPT.”

    Crosby said that, as alleged in the complaint, OpenAI tried to “free ride on the Times' huge investment in journalism by using it to build alternative products without permission or payment.” “It's not fair use even if you think about it,” he added.

    OpenAI accuses New York Times of 'manipulating' ChatGPT in copyright lawsuit

    AI copyright war

    ChatGPT is a free-to-use generative AI trained on billions of texts and codes from the entire internet that existed before 2021. Chatbots have become extremely popular since their launch in November 2022 thanks to their ability to perform a variety of tasks. , writing poetry, etc.

    However, AI companies like OpenAI are facing increasing pressure to use copyrighted material to train large-scale language models. OpenAI and other artificial intelligence companies argue that processing large amounts of publicly available data on the Internet constitutes “fair use” under U.S. copyright law.

    That didn't stop companies from filing lawsuits. In September, about 20 American fiction writers, including John Grisham, George R.R. Martin, and Jodi Pickult, were accused of copyright infringement for using their works in ChatGPT training. He sued OpenAI over the allegations.

    In July, two nonfiction authors filed a similar lawsuit against the company, accusing OpenAI of using their books to train chatbots without their permission. OpenAI is also facing a $3 billion lawsuit over alleged data theft. Last February, Getty Images filed a lawsuit against AI image generation company Stability AI, accusing it of copying 12 million Getty images as training data.

    The New York Times lawsuit comes as OpenAI seeks to enter into agreements with other news publishers to use their content under license. In December, the company reached an agreement worth millions of dollars a year with German publisher Axel Springer that could serve as a template for future deals of a similar nature.

    “We believe the New York Times' lawsuit is without merit. Nevertheless, we look forward to a constructive partnership with the New York Times and respect its long history,” OpenAI said in a statement. I mentioned it on my blog.


    Leave a Reply

    Please enter your comment!
    Please enter your name here