Our discussions with The Latest York Times had gave the impression to be progressing constructively through our last communication on December 19. The negotiations focused on a high-value partnership around real-time display with attribution in ChatGPT, by which The Latest York Times would gain a recent option to connect with their existing and recent readers, and our users would gain access to their reporting. We had explained to The Latest York Times that, like every single source, their content didn’t meaningfully contribute to the training of our existing models and likewise would not be sufficiently impactful for future training. Their lawsuit on December 27—which we learned about by reading The Latest York Times—got here as a surprise and disappointment to us.

Along the best way, that they had mentioned seeing some regurgitation of their content but repeatedly refused to share any examples, despite our commitment to analyze and fix any issues. We’ve demonstrated how seriously we treat this as a priority, reminiscent of in July once we took down a ChatGPT feature immediately after we learned it could reproduce real-time content in unintended ways.

Interestingly, the regurgitations The Latest York Times induced seem like from years-old articles which have proliferated on multiple third–party web sites. It seems they intentionally manipulated prompts, often including lengthy excerpts of articles, with a view to get our model to regurgitate. Even when using such prompts, our models don’t typically behave the best way The Latest York Times insinuates, which suggests they either instructed the model to regurgitate or cherry-picked their examples from many attempts.

Despite their claims, this misuse will not be typical or allowed user activity, and will not be an alternative to The Latest York Times. Regardless, we’re continually making our systems more proof against adversarial attacks to regurgitate training data, and have already made much progress in our recent models.