Is training code and dataset not source?

Are they open?

I’m just challenging your point that ai doesn’t have source.