OpenAI has introduced a formal framework for investigating and publicly reporting cases in which its artificial intelligence systems behave unexpectedly, act without authorization, evade oversight or pursue objectives in ways that conflict with their instructions.
