> Even worse is what the lawsuit calls "regurgitation," which is when OpenAI spits out text that matches Times articles verbatim. The Times provides 100 example

> Even worse is what the lawsuit calls "regurgitation," which is when OpenAI spits out text that matches Times articles verbatim. The Times provides 100 examples of such "regurgitation" in the lawsuit. In its rebuttal, OpenAI said that regurgitation is a "rare bug" that the company is "working to drive to zero." It also claims that the Times "intentionally manipulated prompts" to get this to happen and "cherry-picked their examples from many attempts." So long as an LLM is trained on unlicensed content, this risk can never be driven to zero, because LLMs are not provably correct against regurgitation. OpenAI's rebuttal is marketing that counts on persuading the audience they its engineers can write bug-free software. It's not just Microsoft and OpenAI that are vulnerable to lawsuits when an LLM generates content that may be assessed as a "derivative work" in a court of law — it's any OpenAI user who republishes the derivative work.