OpenAI unveils CSAM classifier that flags AI-generated child-abuse images
OpenAI announced a classifier it says more accurately detects child sexual abuse material (CSAM) and specifically flags AI-generated imagery, reducing human-reviewed false positives by 48% while doubling the hit rate for abusive material. The company is offering vetted organizations free access, providing developer guidance, training law enforcement, and adding new ChatGPT safeguards (a warning plus a 24-hour pause on first offenses); OpenAI says it reports all CSAM to NCMEC.