According to BlockBeats, on September 27, OpenAI launched a new model based on GPT-4O, which is more accurate in detecting harmful text and images.

The new content review model is open to all developers for free through the review API. (Jinshi)

source