At a press briefing in Washington, D.C. on November 16, 2026, OpenAI announced a new image classifier aimed at child sexual abuse material (CSAM). The company says the system can spot both real and AI‑generated abusive images and reduces the number of false positives that reviewers have to handle. OpenAI will give the classifier to selected nonprofits, law‑enforcement agencies and other vetted partners at no cost, and it is also releasing guidance for developers on how to detect this type of content.
The move comes as the National Center for Missing and Exploited Children’s (NCMEC) CyberTipline has been swamped by AI‑related reports. NCMEC logged 400,000 generative‑AI CSAM reports in 2025, up from 67,000 the year before, a surge that officials say could make the tipline’s workload unmanageable. North Carolina Attorney General Jeff Jackson called the increase “explosive” and warned that without new tools the agency could fail its mission.
How the classifier works and what OpenAI added elsewhere
OpenAI’s classifier assigns a score to each image based on visual cues it has not seen before, then directs reviewers toward the most likely abusive content. In early testing with partners, the company reported that the tool flags 48 % fewer images while still catching the same number of abusive ones, and the images it does flag are twice as likely to actually contain abuse. The system also includes a specific flag for AI‑generated CSAM, something OpenAI says is the first of its kind.
Alongside the classifier, OpenAI is updating ChatGPT with warning screens for users who request sexualized content involving children. The first offense triggers a 24‑hour account pause and directs the user to help resources; repeat offenses may lead to permanent account closure. OpenAI says it will report every instance of CSAM it detects to NCMEC.
Why this matters now
Policymakers are pushing for stricter AI safety rules, and several states have already passed laws covering synthetic CSAM. By offering the classifier for free and training investigators, OpenAI hopes to reduce the flood of fake images that can obscure real abuse and to give law‑enforcement a clearer path to act quickly. The approach is still experimental, according to the AGs, but it represents a new model of cooperation between a tech company and child‑protection officials.
Why it matters
For everyday people, the classifier could mean faster removal of illegal content and less chance that harmless images are mistakenly reported. Law‑enforcement agencies get a clearer signal about which images truly need investigation, potentially protecting more real victims. However, the tool’s effectiveness in real‑world deployments is still being evaluated, and it remains limited to organizations that meet OpenAI’s vetting criteria.