In theory, writers whose writing was used as input training data for these LLMs will have a higher likelihood of being misclassified as AI given that the LLM may leverage some unique aspect of their writing style in some outputs.