What changed
On November 16, 2026, OpenAI released a new image classifier capable of detecting potential child sexual abuse material (CSAM), including AI-generated content. This tool is intended to assist organizations in identifying harmful material more effectively. The classifier was developed with input from the National Center for Missing and Exploited Children (NCMEC) and is available through the OpenAI API at no cost to vetted organizations.
What this means in practice
The classifier uses visual indicators to identify CSAM and returns scores to help reviewers prioritize images that may require attention. Early tests suggest it could reduce the number of images needing review by 48% while maintaining similar detection rates as existing tools. Organizations such as fal, invideo AI, and SafetyKit reported high detection rates, with one in four flagged cases leading to bans and 90% detection of cases involving drawings or animations that sexualize children.
Limitations and availability
Access to the classifier is limited to vetted organizations through an intake form. The tool is part of OpenAI’s broader efforts to combat online child sexual abuse, including new product warnings in ChatGPT and practical guidance for developers. The classifier is not yet fully available to the public and requires collaboration with child safety organizations and law enforcement to maximize its impact.
