OpenAI has released a free image classifier that detects child sexual abuse material (CSAM), including AI-generated images, and is adding warnings and 24-hour suspensions in ChatGPT for people who try to find such content. The company announced the tools on 16 November 2026, together with new guidance for developers and AI tools for child-safety police investigators.
Quick Take
- The classifier: detects potential CSAM, including novel images and AI-generated material, and returns risk scores so human reviewers know what to look at first. OpenAI says it is the first broadly available CSAM classifier with a specific signal for whether material may be AI-generated.
- Access: free through the OpenAI API for vetted organizations, via an intake form. Built with help from the National Center for Missing and Exploited Children (NCMEC); access is also shared through the Tech Coalition and ROOST.
- Results: in OpenAI’s evaluation, it would flag 48% fewer images than the current tool while finding roughly the same number of abusive images, and flagged images were about twice as likely to contain abuse.
- In ChatGPT: attempts to find abusive material trigger a warning, removal of the conversation and a 24-hour pause on ChatGPT access. Repeat violations can lead to account closure or reports to authorities.
- For investigators: plugins and tools to review long chat logs and legal process returns, find connections and build timelines, with access managed by program administrators.
In plain English
OpenAI built a tool that can spot images of child sexual abuse, including fake ones made by AI, and it is giving that tool to safety teams at other companies for free. Inside ChatGPT, anyone trying to get this kind of material will now see a warning, lose access for a day and be pointed to help services. Normal users do not need to do anything.
What happened
OpenAI said it trained, evaluated and designed the classifier “with assistance from the National Center for Missing and Exploited Children,” which runs the CyberTipline, the US reporting hub for online child exploitation. Scores from the classifier indicate risk to help prioritize review; they do not on their own confirm that an image is CSAM.
Early access partners reported their own results. In tests by fal on human-reviewed production images, reviewers confirmed 72% of the classifier’s high-confidence flags as novel CSAM. invideo AI said one in four flagged cases led to a ban. SafetyKit said the classifier detected 90% of cases involving drawings or animations that sexualize children.
OpenAI said the ChatGPT warnings were developed with Protect Children, the Lucy Faithfull Foundation, Caoilte Ó Ciardha of the University of Kent and Joel Scanlan of the University of Tasmania, who independently evaluated a trial of deterrent messages.
What’s in the developer guidance
OpenAI also published CSAM guidance for developers, written with input from the Tech Coalition, Thorn, NCMEC and the Internet Watch Foundation. It covers clear rules, user reporting, detection tools, and processes for reviewing and reporting suspected abuse, plus how safety identifiers can link misuse to individual users. OpenAI says the guidance has already led more developers to adopt child-safety tooling and register to report CSAM to NCMEC.
Why it matters
For platforms that host images, especially AI image and video generators, a free classifier with an AI-generated signal fills a gap that hash matching, which only catches known images, cannot cover.
For trust and safety teams, fewer false positives means less exposure of human reviewers to harmful material and faster action on real cases.
For OpenAI, the launch comes as it faces mounting pressure over safety, including the California lawsuit over its AI agents and outside scrutiny of its models through embedded safety assessments.
What’s still open
- How OpenAI decides which organizations count as vetted, and how long approval takes.
- Independent evaluations of the classifier beyond OpenAI’s and its partners’ own tests.
- How the 24-hour suspensions handle false positives, and how users can appeal.
Sources
- OpenAI, “Strengthening the fight against online child sexual abuse with new tools and support,” 16 November 2026
- OpenAI, CSAM Image Classifier model card and Help Center access page
- TechCrunch, “OpenAI releases a new image classifier and ChatGPT warnings to combat CSAM,” 16 November 2026
i10x Perspective
Giving a safety tool away is the right call here. CSAM detection only works if smaller platforms can afford it, and many AI image startups cannot build this themselves. The AI-generated signal matters most: synthetic abuse material is the problem generative AI created, and the industry needs shared defenses against it. More in our Policy & Safety section.






