“The New York Times ‘hacked’ ChatGPT for fake evidence”


OpenAI claims that The New York Times exploited a bug in its AI chatbot to gather evidence for the lawsuit.

In the ongoing lawsuit between the two, The New York Times was able to prove that ChatGPT completely copied paragraphs of text from newspaper articles. That evidence added weight to their argument that the chatbot violates copyright law. According to OpenAI, that piece of evidence is invalid: the New York Times allegedly used a bug to influence the results of the chatbot. Certain misleading prompts bypassed the security system, copying text directly from the Internet.

OpenAI then released one statement released claiming that The New York Times hired someone to ‘hack’ the AI ​​system. The newspaper is said to have made ‘great efforts’ to have ChatGPT reproduce verbatim passages from their articles. It would take tens of thousands of attempts to get to that result. In addition, OpenAI claims that it only concerns limited parts of articles that can already be read for free in other places on the internet.

The lawyer for The New York Times, Ian Crosby, reacts with surprise to the accusations in a note to The Register. According to him, they used ChatGPT the way everyone else does and found the evidence. Additionally, this response shows that OpenAI is tracking users’ queries, which they actually claim not to do.

ChatGPT lawsuit

The New York Times does not agree with the way ChatGPT gathers its knowledge. The chatbot scrapes the entire internet and then uses that information to reproduce similar answers. According to OpenAI, this is done under a ‘fair use’ license and the chatbot would not copy literal pieces of text from the internet. The American newspaper does not agree with that definition and is therefore filing a lawsuit. She claims that ChatGPT does provide literally copied answers, including from the online material of the American newspaper.

OpenAI has often come under fire for their use of copyrighted material. This is how they came into the waters of the GDPR legislation in Italyand dragged various authors them in court on the grounds of copyright.


Featured article

openai

OpenAI admits: “Training AI with only royalty-free content is not possible”