OpenAI unveiled a classifier it says can detect child sexual abuse material, including whether imagery is AI-generated, and will give vetted organizations free access alongside developer guidance and law enforcement training. The move comes as NCMEC's CyberTipline saw generative AI-related reports surge from 67,000 in 2024 to 400,000 in 2025. OpenAI characterized the effort as still essentially an experiment in pre-harm prevention.
NCMEC
美国失踪与受虐儿童中心
US nonprofit operating the missing and exploited children hotline and partnering with tech firms on CSAM response.
Curated Intelligence
The Verge AIT2·Industry
62 pts
TechCrunch AIT2·Models
62 pts
OpenAI released an image classifier that detects CSAM, including AI-generated material, offering free API access to select companies and co-developing it with NCMEC. ChatGPT will also issue warnings and temporary blocks for content sexualizing children, alongside a developer guideline and law-enforcement tooling. Early partner testing cut flagged images by 48% while catching the same number of abusive images, easing reviewer workload.