To manage its risks, “superintelligent” artificial intelligence should be governed by a body similar to the International Atomic Energy Agency, the lab’s leadership said in note on its website.