What The Hell America!

A federal judge on Wednesday rejected OpenAI’s request to toss out a copyright lawsuit from The New York Times that alleges that the tech company exploited the paper’s content without permission or payment.

In an order allowing the lawsuit to go forward, Judge Sidney Stein, of the Southern District of New York, narrowed the scope of the lawsuit, but allowed the case’s main copyright infringement claims to go forward.

Stein did not immediately release an opinion but promised one would come “expeditiously.”

The decision is a victory for the newspaper, which has joined forces with other publishers, including The New York Daily News and the Center for Investigative Reporting, to challenge the way OpenAI collected vast amounts of data from the web in order to train its popular artificial intelligence service ChatGPT.

Lawyers for The Times believe that the paper’s articles are one of the biggest sources of copyrighted text OpenAI used to build ChatGPT into the premiere AI chatbot, and allege that OpenAI violated copyright laws in its siphoning of the newspaper’s journalism.

The Times and OpenAI did not immediately return requests for comment.

OpenAI leaders have argued that the company’s mass data scraping, including articles from The Times, is protected under a legal doctrine known as “fair use.”

It allows for material to be reused without permission in certain instances, including for research, teaching and commentary.

The judge’s ruling means that the suit can now proceed to trial, but a trial date has not been set. Evidence-gathering, including depositions with executives on both sides of the case, are expected to happen confidentially, along with public pre-trial hearings to settle disputes over evidence and other matters.

The legal fight between one of the world’s most influential news outlets and Silicon Valley’s leading AI company puts a lot on the line for both the news industry and the future of AI tools.

For publishers, the fear is that powerful chatbots that can quickly summarize news articles when queried will mean that readers visit news websites less often, prompting a drop in advertising that could affect the industry’s bottom line.

And while the suit only names OpenAI and its financial backer, Microsoft, other AI companies also scrape the web for content to train their models. For the most part, the AI industry has followed OpenAI’s lead when it comes to training chatbot and other AI services, operating under the premise that processing data found on the open web into chatbot answers is legally protected by copyright law.

But the law remains unsettled on the matter.

Courts have said fair use of a copyrighted work must generate something new that is “transformative,” or comments on or refers back to an original work. The Times argues that this does not apply to how OpenAI reproduces the paper’s original reporting.

Another part of the legal analysis will involve an idea known as “market substitution,” referring to whether chatbot answers are a substitute for, say, reading The Times website, or if chatbots and newspapers operate in different marketplaces.

In a hearing in New York in January, lawyers for the publishers argued that when ChatGPT was asked questions about subjects covered by The Times, the chatbot regurgitated articles verbatim.

Yet OpenAI’s legal team shot back that the news outlets appeared to have specifically manipulated prompts to essentially force the chatbot to spit out large chunks lifted from the paper’s website. This is not how most people interact with the service, nor how the chatbot is designed to operate, said Joseph Gratz, a lawyer for OpenAI.

“This isn’t a document retrieval system. It is a large language mode,” Gratz said.