OpenAI flags 6 new examples of 'concerning' AI behaviour

OpenAI has disclosed six reports of "unexpected or concerning" behaviour in artificial-intelligence models as the debate on AI safety becomes increasingly heated.