They literally do not pass the criteria. LLMs use the entirety of a copyrighted work for their training, which fails the “amount and substantiality” factor. By their very nature, LLMs would significantly devalue the work of every artist, author, journalist, and publishing organization, on an industry-wide scale, which fails the “Effect upon work’s value” factor.
Those two alone would be enough for any sane judge to rule that training LLMs would not qualify as fair use, but then you also have OpenAI and other commercial AI companies offering the use of these models for commercial, for-profit purposes, which also fails the “Purpose and character of the use” factor. You could maybe argue that training LLMs is transformative, but the commercial, widespread nature of this infringement would weigh heavily against that. So that’s at least two, and arguably three out of four factors where it falls short.