OpenAI trained o1 and o3 to 'think' about its safety policy | TechCrunchOpenAI has launched o3, a new AI reasoning model family that incorporates a new safety training paradigm to enhance alignment with human values.