OpenAI has introduced an image classifier designed to identify child sexual abuse material and indicate whether it was generated by AI. The company is offering free access to vetted organizations, alongside training for law enforcement and guidance for developers building detection systems.

The release addresses a triage problem at the National Center for Missing and Exploited Children. Its CyberTipline received 400,000 reports related to generative AI in 2025, up from 67,000 in 2024. Synthetic material is illegal under US federal policy and covered by laws in 48 states, but a flood of generated images can delay investigations involving children in immediate danger.

OpenAI says the classifier produces 48% fewer flagged images than its previous approach while making each flag twice as likely to contain abusive material. It can also recognize synthetic content from other systems, including open models and “nudifying” apps. Those performance figures are company claims; the report does not provide an independent benchmark or error rates across different image sources.

ChatGPT will also show a warning and impose a 24-hour pause after a first attempt to generate prohibited material, while directing the user to support resources. North Carolina Attorney General Jeff Jackson called the preventive partnership experimental. Its usefulness will depend on detection accuracy, careful access controls and whether investigators can prioritize real victims without treating synthetic abuse as harmless.