Why is this a problem?

If the courts are using ai we should know why and how.

There should be a human reviewing it.

The judge should be making their own decision that shouldn't be based on automated tools.

   [The judge] said he welcomed litigants’ use of AI in general  and even used AI tools to help prepare the decision. But hiding  a set  of instructions is “evidence of its malicious purpose,” he wrote.

You cannot, for example, submit a printed filing with temperature sensitive ink which alters the text when scanned on a flatbed scanner, and then expect to not by sanctioned because you claim you're auditing whether or not the court used optical character recognition.