OpenAI Unveils Tool to Help Investigators Combat AI-Generated Child Sexual Abuse Material
· GlobeWire Daily Editorial Desk · Source: www.theverge.com

Image: Lauren Feiner · Photo source · Resized and letterboxed to 16:9; no AI generation
Addressing the CSAM Backlog
OpenAI has introduced a new classifier designed to assist investigators in identifying child sexual abuse material (CSAM), including content generated by artificial intelligence. The tool aims to reduce the volume of false positive reports, allowing organizations like the National Center for Missing and Exploited Children (NCMEC) to focus resources on urgent cases.
The Scale of the Problem
In 2025, NCMEC’s CyberTipline reported receiving 400,000 reports related to generative AI, a significant increase from 67,000 in 2024. Experts warn this surge threatens to overwhelm existing triage systems. North Carolina Attorney General Jeff Jackson characterized this influx as "explosive," noting that it risks undermining the ability of law enforcement to protect children from active harm.
Key Features of the New Classifier
- Reduction in False Positives: OpenAI claims the classifier flags 48 percent fewer images than previous methods, with a higher accuracy rate for abusive material.
- AI Detection: It is the first widely available classifier that specifically flags content as AI-generated, helping investigators distinguish between fictional and real-child imagery.
- Educational Support: The tool is being offered to vetted organizations, accompanied by guidance on how to train law enforcement personnel.
User Intervention and Prevention
OpenAI is also implementing new protocols for its own platforms. If a user attempts to generate CSAM through ChatGPT, they will receive a warning and be directed to support resources. A first-time violation will result in a 24-hour account suspension, with further warnings that repeated offenses may lead to permanent closure.
Verification and Limitations
While OpenAI emphasizes this is a collaborative approach to harm prevention, officials like Jeff Jackson describe the partnership as an "experiment." Interested parties can verify the efficacy and technical integration of these tools through documentation provided to vetted organizations or official announcements from NCMEC regarding their triage workflows.
AI-assisted reporting; source claims are attributed and have not been independently verified.



