Did that marketing bot actually cause harm though, or was it just a bit off? Sometimes chasing perfect data costs more than just letting a few old specs slide. Unless it's breaking legal rules or really confusing customers, a slightly messy dataset might not be worth the crazy cleanup effort. Perfect information is impossible anyway. How much time are we supposed to spend scrubbing every old file?
Ever seen an AI get stuck on old specs? My friend's marketing bot kept writing press releases for a product name they changed two years ago. They had to scrub all the old project briefs, which is exactly what lily_murray68 is asking about. Wasted a full week because the training data was a messy mix of old and new.