There are different types of risk associated with AI (eg propaganda en mass, deep fakes, breaking codes, etc). I wonder which ones they were working on and which ones are societal problems vs their problems.