OpenAI Announces Child-Safety Tools and ChatGPT Warnings
OpenAI has announced a set of child-safety measures that combine an image classifier for detecting potential child sexual abuse material, new ChatGPT warnings and temporary suspensions, developer guidance, and tools for investigators. The company says early partner testing found the classifier reduced images sent for human review by 48% while identifying roughly the same amount of abusive material as its existing system.
The classifier is designed to flag visual indicators in images, including material generated by artificial intelligence and images not previously identified. It assigns risk scores to help human reviewers prioritize cases; it is not intended to decide automatically whether an image is illegal. OpenAI said images flagged by the new system were twice as likely to contain abusive material as images flagged by its prior system. The reported performance figures come from company-described early testing, and the reports do not provide independent validation.
OpenAI is offering free API access to the classifier for approved organizations and said it is working with the Tech Coalition and ROOST to expand access. The company developed the tool with assistance from the...