Then what I’d like to see eventually is an equivalent of a CERN for AI safety that does research into that – but internationally. And then maybe there’s some kind of equivalent one day of the IAEA, which actually audits these things. (source)