Reading through it again, it's interesting how it's relevant not only to the topic of money laundering but also to the question of whether and to what extent AI-generated output is "colored" by the copyright of the training data.