NYT vs. OpenAI, Chapter XXVIII

Credit: DALL-E

So as not to mess up the launch of its GPT Store, OpenAI decided to respond to The New York Times’ lawsuit the day before. It did so via blog post, which may not have been the most considerate choice, but no less than when the Times filed its legal complaint without a heads-up. (OpenAI says in the post that it first became aware of the lawsuit by reading the Times’ story about it.)

As I predicted, OpenAI draws a distinction between normal responses and regurgitation, which amounts to spitting back near-exact copies of training data — the primary evidence the Times points to in its lawsuit. OpenAI says regurgitation is not a normal use case and that it’s taken steps to stamp it out from ChatGPT’s responses. It accuses the Times of using essentially cherry-picked examples of non-representative use cases.

On the subject of whether OpenAI had the right to hoover up the Times archive in the first place, it simply asserts the act as fair use, and leaves it at that. That’s obviously the only stance the company could take, given that its models entirely depend on crawling the web for training data.

Whether or not the courts agree — or take the perspective that the novel use of the data to create products that potentially negate the need to visit The New York Times or any other news website — remains to be seen. Many are predicting that the U.S. Supreme Court will have the final word.

Finally, OpenAI addresses the apparent contradiction in its stance on fair use and the deals it’s signed with Axel Springer and the Associated Press. These are partnerships created to access proprietary archives and “display real-time content with attribution,” OpenAI says, not an admission that it didn’t have the right to harvest their data in the first place.

Subscribe now

The View From the Sidelines

It’s fair to say that every media company has a stake in the outcome of this lawsuit. However, if you’re an executive at another publisher watching from the s The questions is: What do you do now?

Broadly, I see three choices:

Join in: Sue OpenAI yourself, throwing your lot in with the Times, ostensibly increasing the chance of a redefinition of copyright law for AI.

Cut a deal: Connect with OpenAI on a partnership similar to Axel Springer’s. While the amount of cash that OpenAI appears to be offering for these deals is relatively low (a report in The Information pegged it between $1-5 million), some money is better than no money.

Wait it out: Stay on the sidelines and see what the final decision is, prepping for either outcome. If they settle, that’s essentially a victory for OpenAI since it restores status quo, and you’re back to No. 2. Certainly, if you want to keep your options open, this is the smartest way to go, and likely where most observers are landing.

There may be a fourth option, though: collective action. Shortly after the initial lawsuit dropped, Francesco Marconi, the CEO of AI media lab AppliedXL, wrote an opinion piece for the Wall Street Journal, arguing for a middle ground between the AI and journalism camps. “For AI to flourish, journalism must thrive,” he wrote, explaining that it was in the interest of companies like OpenAI to support the original work that reporters and editors create.

After OpenAI responded, Marconi wrote on LinkedIn that the parties involved should work toward a “new paradigm of IP management,” one that determines a “fair market price” for the data that AI models need to give the best information.

Such cooperation would certainly be ideal. But given how prevalent the use of public internet data is in training LLMs, it’s hard to see AI creators being excited about coming to the negotiating table — unless of course they get sued. Maybe that was the point all along.

Join us in New York on February 1! We are planning our first meetup in Manhattan and we’d love to meet you! Sign up to our Meetup Group here and we’ll send you the details shortly.

Ready to start using AI like a pro?


Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.