I'm not sure that's a sufficient argument.

1. Laws can change.

2. Blackbox models can provide specific reasons, even if they need to hallucinate them.