They're blaming the court. While there is an order, it is happening in response to massive, blatant, and continued I.P. infringement. Anyone doing that knows they'll be in court at some point. Might have a "duty to preserve" all kinds of data. If they keep at it, then they are prioritizing their gains over any losses it creates.

In short: OpenAI's business practices caused this. They wouldn't have been sued if using legal data. They might still not have an order like this if more open about their training, like Allen Institute.