AI Control: Improving Safety Despite Intentional Subversion.

Ryan Greenblatt, Buck Shlegeris, Kshitij Sachan, Fabien Roger

Browse the full ICML paper archive.