According to CoinMeta, OpenAI has updated its model specifications, restructured the guidelines around five core values: aligning with human intentions, minimizing harm, honesty and transparency, human autonomy, and humanistic care. It also explores new training methods to help the model better understand and adhere to relevant behavioral guidelines.