Written by 6:35 pm ChatGPT, OpenAI

### OpenAI Report: ChatGPT Duped into Replicating The New York Times’ Material

OpenAI insists training AI models on copyrighted data is fair use.

OpenAI has publicly addressed a trademark lawsuit from The New York Times, dismissing the issue as “insignificant” and expressing a desire to maintain a relationship with the news outlet.

In a statement on their website, OpenAI suggested that the Times was not providing a complete picture of the situation. They specifically refuted allegations that their ChatGPT AI tool replicated Times articles verbatim, contending that the Times had manipulated the prompts to include repeated segments of their articles. OpenAI stated that despite using such prompts, their AI models did not consistently produce the outcomes implied by The New York Times, implying that the news outlet either directed the model to repeat content or selectively chose instances from multiple outputs.

OpenAI asserted that they had made efforts to minimize duplication from their extensive language models and claimed that the Times had declined to share examples of the alleged replication prior to initiating legal action. They clarified that the identified examples seemed to originate from older articles that had been widely disseminated on various third-party platforms. The company acknowledged removing a ChatGPT feature called Browse that inadvertently reproduced articles.

Despite these developments, OpenAI reiterated its longstanding position that for AI systems to evolve and tackle new challenges, they require access to a vast array of human knowledge. They emphasized their respect for copyright laws and the provision of opt-outs for training data inclusion but argued that training AI models with internet data falls within fair use guidelines permitting the repurposing of copyrighted material. OpenAI announced that website owners could block their web crawler from accessing their content starting in August 2023, nearly a year after the ChatGPT launch.

The company presented a similar argument to the UK House of Lords, asserting that AI systems like ChatGPT must incorporate copyrighted material to comprehensively represent human knowledge and experiences.

OpenAI expressed a willingness to engage in further discussions with The New York Times to establish a partnership akin to those formed with Axel Springer and The Associated Press. They expressed optimism about fostering a collaborative relationship with The New York Times while acknowledging and respecting its rich journalistic history.

Visited 1 times, 1 visit(s) today
Last modified: January 9, 2024
Close Search Window