With complexity rising and certainty elusive, the most unshakable leaders are not the ones with all the answers; they are the ones who know where to lead from.

Chain-of-thought monitorability could improve generative AI safety by assessing how models come to their conclusions and ...