OpenAI and the New York Times: a battle of giants
The dispute between the New York Times and OpenAI is heating up. The accusation is serious: OpenAI allegedly concealed evidence in a trial regarding copyrights involving ChatGPT. The Times and the Daily News claim that OpenAI lied about its ability to search chat data and training sets that would include copyrighted works. This lawsuit has been ongoing for two years and revolves around the accusation that OpenAI used content from the Times to train its AI models, reproducing this material in responses generated by ChatGPT.
OpenAI, in contrast, argues that it has no way to search its own training corpus. They say it would be technically complicated and would raise user privacy issues, as the logs would need to be processed and anonymized. However, according to a testimony from Vinnie Monaco, OpenAI's data privacy engineer, the company had already conducted internal searches to check for any protected journalistic works in its training corpus. Furthermore, even before the lawsuit, OpenAI had accumulated a database of about 78 million ChatGPT conversations, used internally to measure how much they were infringing copyrights.
Revelations that complicate the scenario
These revelations are a goldmine for the Times. They requested a sample of 120 million chat logs, but OpenAI negotiated to reduce this to 20 million. When they finally delivered the sample last December, it was so filled with redactions that the court deemed it "unusable." The accusation is that OpenAI deleted billions of ChatGPT outputs after the lawsuit began, violating court orders to preserve data, and that it replaced millions of logs in the requested sample.
If OpenAI believed that copying journalism from our clients was fair and legal, it would not have hidden the truth, said Ian B. Crosby, lead attorney for the claimants. Now, the Times and the Daily News want the judge to penalize OpenAI for allegedly concealing evidence and obstructing the discovery process. They are asking the court to prevent OpenAI from using the sample of 20 million logs as evidence, claiming it is unreliable, and to accept as a fact that the ChatGPT logs would show significant regurgitation of the claimants' content.
Comments (0)
Comments are moderated and if they violate our Terms and Conditions of use, the comment will be deleted. Persistence in violation will result in a ban of your account.