The original research paper on IBM’s AI Risk Atlas defines a taxonomy of 40 AI risks. IBM then expanded the online Atlas, and the current IBM documentation lists 100 named risks. Where does the 60 number come from for IBM's list of AI risks?