AI labs tend to use "system cards" to describe their evaluation and safety research processes.

They used to be more about the training process itself, but that's increasingly secretive these days.