A harm severity scale across 10 categories based on the CSET AI harm taxonomy with quantifiable criteria to support objective assessment. Scores reflect impacts ranging from 1 (‘negligible’) to 5 ('catastrophic')

‍

‍