Free for vetted organizations, AI-generated images included
OpenAI on Nov. 16 (early Nov. 17 Japan time) announced that it will provide a new tool to help detect child sexual abuse material (CSAM) to vetted organizations free of charge. CSAM refers to images, videos and other content that record the sexual abuse of children. According to OpenAI, the detection also covers AI-generated images. The announcement came from the company's official newsroom account on X, "OpenAI Newsroom."
We're releasing a free new tool to vetted organizations to help detect child sexual abuse material, including AI-generated images. We're also adding ChatGPT warnings, publishing guidance for developers, and supporting child safety investigators. We developed the classifier with assistance from the National Center for Missing & Exploited Children. We're working with the Tech Coalition and ROOST to help make it available to more organizations. https://openai.com/index/strengthening-online-child-safety/
Classifier developed with NCMEC
OpenAI said it developed the tool's classifier (an AI system that sorts data by type) with assistance from the U.S. National Center for Missing & Exploited Children (NCMEC). It said it is working with two organizations, the Tech Coalition and ROOST, to expand the range of organizations that can use it.
Warnings in ChatGPT and guidance for developers
Besides the detection tool, OpenAI listed three other steps. They are adding warnings to ChatGPT, publishing guidance for developers and supporting child safety investigators. The post does not say in what situations the warnings will appear or what the guidance and support will specifically involve.
Vetting criteria and timing not given
The post describes the recipients of the tool as "vetted organizations" but does not give the vetting criteria or how to apply. It also does not state when the tool becomes available, its detection accuracy or the tool's name.
The post links to an official OpenAI blog post, but when ORIENS tried to retrieve it, access was denied (HTTP 403), and the blog's content could not be confirmed. This article is based solely on the text of the official X post.






