OpenAI launches image classifier to detect child sexual abuse material

What happened
OpenAI has launched a new image classifier designed to detect child sexual abuse material (CSAM), including content generated by AI. The company also added warnings and temporary account blocks in ChatGPT for users who attempt to generate CSAM content. These measures were announced at a press briefing in Washington, DC, and are being offered to vetted organizations and law enforcement. OpenAI also released new developer guidelines and training resources for investigators.
Key details
- The classifier is available for free to vetted organizations.
- NCMEC's CyberTipline received 400,000 generative AI-related CSAM reports in 2025, up from 67,000 in 2024.
- The new classifier flags 48% fewer false positives and is twice as likely to identify abusive material.
- ChatGPT users who attempt to generate CSAM receive a warning and a 24-hour account pause on first offense.
- OpenAI provides guidance and training to law enforcement and developers to help identify and report CSAM.
Why it matters
Child safety investigators and organizations now have access to a tool that can more accurately detect both real and AI-generated CSAM, helping them manage a rapidly increasing number of reports. ChatGPT users face immediate consequences and support interventions if they attempt to generate harmful content. These changes aim to reduce the burden on reporting systems and assist law enforcement in responding more effectively to child safety threats.

