“OpenAI did not have a unique data advantage for training its models, as it primarily used publicly available web-scraped data.”