A company that tests the security of A.I programs found that 2 Chinese ones were willing to devise ways of launching terrorist attacks on London. Are others any safer? Not one is saying.