OpenAI has introduced a new image-classification tool that analyzes visual indicators to detect previously undiscovered child sexual abuse material. The system assigns scores that help moderators determine which images to review first.
OpenAI said this is the first widely available classifier to provide a separate signal for such material generated by artificial intelligence. Verified organizations will be able to use it free of charge.
ChatGPT blocks access for 24 hours

ChatGPT will also display warnings interrupting attempts to search for such content. Users will be told that these materials cause harm, even when they are generated by artificial intelligence.
The flagged conversation will be deleted, and the user's access to ChatGPT will be suspended for 24 hours; during that time, the user will be offered help from specialists. Repeated violations may lead to longer restrictions, account deletion or the transfer of information to the authorities.
OpenAI said it developed this approach with organizations working to combat violence against children and with academic advisers.