Hmmm, so the NSF now requires a paragraph of text in grant applications and annual reports where the PI claims to have provided mentoring? That's better than nothing, but I have to be skeptical about its having any real impact. There's an issue of culture that's really difficult to change, and while I wish I had something constructive to propose, I'm drawing a blank.

How about changing the standard acknowledgment of NIH funding in publications to require a precise description of the contribution of the grad student/postdoc? When PI's have one of those postdocs who is a pair of hands in the lab but doesn't write papers (the large majority, in my experience) let them say so. That has a slightly higher barrier to wishful thinking than does some boilerplate about "training in research ethics".