OpenAI rolled out a broad child-safety package centered on a new CSAM image classifier, plus ChatGPT interventions, developer guidance and investigative support for child-safety law-enforcement units.
Early tests showed the classifier could flag 48% fewer images while finding roughly the same number of abusive images; high-confidence flags were confirmed as novel CSAM in 72% of fal’s reviewed cases.
The tool, built with NCMEC input, is available at no cost through the OpenAI API to vetted organizations and is designed to detect potential AI-generated CSAM as well as previously unseen abusive images.
ChatGPT will now warn users seeking abusive material, delete the flagged conversation and pause access for 24 hours; repeated violations can bring longer restrictions, account closure or reports to authorities.
OpenAI also published developer guidance with industry partners and introduced training and tools for investigators, extending its child-safety push beyond detection toward prevention, reporting and casework.