[Terr_]

The authors' focus here seems to be on considering AI as something with its own goals and intent, however there's another category of risk which is much closer: What humans will do to humans using AI as a tool.

Consider the difference between "I worry inventing nuclear bombs will make us extinct" versus "I worry inventing the Hypermind from the book Don't Create The Torment Nexus will make us extinct." The former doesn't have agency.