I know what a honeypot is, but the question is how the know the scraped data was actually used to train llms. I wondered whether they discovered or verified that by getting the llm to regurgitate content from the honeypot.