Why not ?
If they have something that is actually working, then why not apply it to another domain where it could be useful ?
It's not because it's called AI that it is automatically useless, after all. I'm sure Nippon scientists are going to make it work and, when they do, it will certainly be beneficial for all satellites. Corporate secrets (or National Security) aside, I'm sure every satellite managing company would like to have advanced warning of when things might go wrong.