STEVE INSKEEP, HOST:
If you should ask ChatGPT a question, the answer from that chatbot on the internet may include information first reported by The New York Times. The Times says it never gave ChatGPT or its maker, OpenAI, permission to do this. So the newspaper and other publishers sued OpenAI for copyright infringement. The case may depend on a hearing that happened yesterday in New York, which NPR's Bobby Allyn was covering. Hey, there, Bobby.
BOBBY ALLYN, BYLINE: Good morning, Steve.
INSKEEP: What was it like in court yesterday?
ALLYN: Well, it was pretty contentious. Both sides were arguing before a federal judge about whether this case should move to trial or be dismissed. And the judge didn't rule right away. He said he will rule soon. And as you mentioned, this is a copyright case. So pretty wonky. The details get pretty dense, but it boils down to this - was it legal for OpenAI to feed ChatGPT The Times' digital archive as it was scanning the entire internet? And The Times says, absolutely not. Lawyers for the paper say they never approved it, they never paid for it and they say they have examples of the chatbot reproducing verbatim Times' articles.
INSKEEP: Yeah. They write the first few words of the article in a prompt, as they say, and the entire article comes back, except with that attribution to The Times or paying The Times. So how did ChatGPT's owner respond?
ALLYN: Yeah. OpenAI's legal team told the judge that ChatGPT isn't a document retrieval system, it's a large language model, meaning it's not made to spit out entire Times articles and that The Times was only able to get ChatGPT to do this after asking it thousands of different questions.
INSKEEP: They're not denying that they used New York Times articles, but they're saying they didn't use the exact thing always, and that it's legal. How can that be?
ALLYN: Well, OpenAI says the mass scraping of texts from all corners of the internet, in order to give ChatGPT an enormous trove of data was done under something called the Fair Use Doctrine. And it's a legal principle that says a copyrighted work can be used without permission for things like research, education, commentary, journalism, like we're doing right here...
INSKEEP: Yeah.
ALLYN: ...But it cannot be used in a way where it ends up competing with the original and it has to be transformative, meaning it's got to be made into something new. OpenAI says fair use is a shield that should protect it from this and all copyright lawsuits. But The Times says no way. We should note here that other publishers, including The New York Daily News and the Center For Investigative Reporting have also joined this litigation with The Times, and Microsoft is also being sued here since it incorporated ChatGPT into its Bing search engine.
INSKEEP: OK. You've given us an idea of the arguments. What's at stake here both for publishers and for AI makers?
ALLYN: Yeah. Quite a lot. The news business has been threatened by AI tools, particularly chatbots and so-called answer engines that extract information from articles and synthesize it at lightning speed. And of course, the fears for the news industry is that people will no longer visit publishers' websites like nytimes.com. If you can get it faster from an AI tool like ChatGPT. And of course, could quickly hurt a publisher's bottom line. And for AI companies, Steve, this is huge because it will be the first test of whether the mass scraping these firms did was legal or not. The whole industry has basically been operating under this principle that fair use protects them. And if that is wrong, well, it could really open the floodgates to more suits coming after popular AI chatbots, and there are many out there. Now, in the Times case, Steve, the most dramatic outcome here is the judge ordering that OpenAI destroy its entire data set.
INSKEEP: Wow.
ALLYN: The Times has asked for that.
INSKEEP: NPR's Bobby Allyn, Thanks.
ALLYN: Thanks, Steve. Transcript provided by NPR, Copyright NPR.
NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.
