Because the language models of artificial intelligences have to be trained to give the expected results, the information that is fed to them has to come from somewhere. Since ChatGPT became popular, there have been cases of it being fed copyrighted material. A few days ago, the New York Times published that the process had begun for a complaint against Microsoft y OpenAI for feeding ChatGPT with your articles, a use case that is not included anywhere for free or commercial use.
And it's not just that Copy and paste the items and give them to feed your language model. The lawsuit alleges that they have circumvented the portal's paywall, and attributed disinformation to them that they never wrote.
New York Times vs OpenAI for misuse of its articles
New York Times argues that, to pay its workers, who obtain the information and articles they publish, should put up a paywall. And in their editions, they are aware of copyright, so the Times's terms of service limit copying and use of any published material, and they can be selective about how they license their stories and how they are published.
The tools developed by OpenAI violate all of this, providing content from the New York Times without your permission or authorization. With their tools, they have obtained 16 million unique records from sites published by The New York Times. This places the Times as the third most referenced source by ChatGPT, behind the Wikipedia and a US patent database. Therefore, OpenAI would no longer reveal as many details of the data used for training the latest versions of GPT.
The lawsuit alleges that it is easy to get ChatGPT to offer content behind the paywall. It's okay to ask for an article, ChatGPT delivers the first paragraph, and then it's okay to ask for subsequent paragraphs. Or at least it was possible until very recently. As generative Artificial Intelligence and language models are not fully regulated, such a case could generate some jurisprudence and affect the future of Artificial Intelligence.
