Comment by patwolf

That last one seems like a pretty big deal though. It's not just how you tested, but "other critical information" about the model.

I imagine the government can deem any AI to be a "serious risk" and prevent it from being made public.