On AI Risk › Joe Carlsmith Joe Carlsmith on AI risk Joe Carlsmith: Researcher (Claude constitution) at Anthropic; Senior advisor at Coefficient Giving (Open Philanthropy). This page collects 2 quotes from 2026, newest first, on pause (2), extinction risk (1), loss of control (1), race dynamics (1).
Quotes (2), newest first “Safety progress takes time. If you don’t have enough time, at each stage of AI development, to ensure the AIs that you build in the next stage won’t destroy humanity, then you’ll fail, and humanity will be destroyed. Automated alignment research helps a lot here, and that’s a core reason I think it’s so important (human alignment researchers are too scarce and slow). But even if AIs are doing most or all of the alignment work, it still matters how much time they have.” Joe Carlsmith · March 19, 2026
“Overall, my current view is that despite these uncertainties and trade-offs, we should try hard to put ourselves in a position to steer or restrain AI development when doing so is necessary for safety; that even limited or temporary success in this respect can make a significant positive difference; and that more thoroughgoing success remains, at least, a live possibility.” Joe Carlsmith · March 19, 2026
Colleagues in this collection People who have worked at the same organization, at any time.