US Treasury Secretary Scott Bessent says AI labs should slow development if they fear losing control. He made the remarks in an Axios Show interview published on October 3.
Bessent describes current model reviews as voluntary. He says the government retains the option to intervene if concerning developments at a lab require it.
He places responsibility for safe development on the labs and calls for greater attention to resilience and defense. He does not announce a new binding review regime.
His proposed communication channel with China was already known. The new remarks on development pace are distinct; his account does not establish the effectiveness of the reviews.