OpenAI Releases Free CSAM Detection Tool for Vetted Groups
OpenAI said in a blog post on Nov. 16 that it is releasing a new image classifier to detect child sexual abuse material, including AI-generated images. The company also announced new warnings in ChatGPT, guidance for developers and tools for child-safety investigators.
The classifier was built with help from the National Center for Missing and Exploited Children. Vetted organizations can get it through the OpenAI API at no cost, and they can request access through an intake form. OpenAI said it is the first widely available tool of its kind with a specific signal for whether an image may be AI-generated. It gives reviewers scores so they can decide which images to check first.
OpenAI shared early numbers. In its own test on user-uploaded images, it estimated the new tool would flag 48% fewer images than its current protections while finding about the same number of abusive ones. Flagged images were about twice as likely to contain abuse. Partner company fal said reviewers confirmed 72% of high-confidence flags as new abusive material. SafetyKit said the tool caught 90% of cases involving drawings or animations.
For people who use ChatGPT, the change is a new warning. If someone tries to find abusive material, the warning says sexualizing children does harm, even when the content is AI-generated. OpenAI said it will remove the conversation, pause the person's access for 24 hours and point them to specialist support. Repeated violations can bring longer limits, account closure or a report to authorities. Chatbots can also surface real-world danger, as in a case where a chatbot flagged planned crime in Budapest.
OpenAI also published guidance for developers, written with input from the Tech Coalition, Thorn, NCMEC and the Internet Watch Foundation. For police, it is offering training and tools, starting with one that helps investigators review long chat logs and build timelines. Access depends on agency authorization.
Some things are still unclear. The post does not say when the ChatGPT warnings reach all users or how many organizations will get the classifier. The performance figures come from OpenAI and its early partners. The post does not mention an outside check, a topic we cover in our piece on independent AI evaluators.
Sources
Cite this
OpenAI Releases Free CSAM Detection Tool for Vetted Groups, New Mind Museum, November 16, 2026. https://newmindmuseum.com/safety/2026/11/openai-csam-classifier-child-safety-tools/
New Mind Museum. (2026, November 16). OpenAI Releases Free CSAM Detection Tool for Vetted Groups. https://newmindmuseum.com/safety/2026/11/openai-csam-classifier-child-safety-tools/
“OpenAI Releases Free CSAM Detection Tool for Vetted Groups.” New Mind Museum, 16 Nov. 2026, https://newmindmuseum.com/safety/2026/11/openai-csam-classifier-child-safety-tools/.
@misc{new-mind-museum-2026-openai-releases-free-csam-detection-tool,
author = {{New Mind Museum}},
title = {OpenAI Releases Free CSAM Detection Tool for Vetted Groups},
year = {2026},
url = {https://newmindmuseum.com/safety/2026/11/openai-csam-classifier-child-safety-tools/}
}Free to quote and cite with a link to this page.