OpenAI Launches Free CSAM Classifier Built With NCMEC That Halves False Positives

OpenAI on Sunday released a suite of child safety tools, including a new image classifier designed to detect child sexual abuse material (CSAM) — including AI-generated imagery — and a warning system inside ChatGPT that temporarily suspends users who attempt to generate or access such content.
The classifier, co-developed with the National Center for Missing and Exploited Children (NCMEC), is being made available free of charge through OpenAI's API to select companies working in content safety. It works by analyzing visual signals in images it has never seen before and assigning a risk score to help human reviewers prioritize their queues.
In early partner testing, the classifier flagged 48% fewer images for review while identifying the same number of abusive images. Images it did flag were twice as likely to contain abuse material — a meaningful improvement in reviewer efficiency at a time when the volume of AI-generated CSAM is growing rapidly.
The Tech Coalition and ROOST are helping distribute API access to child-protection organizations.
ChatGPT Warnings and Suspensions
Alongside the classifier, OpenAI is deploying new deterrent measures inside ChatGPT. Users who attempt to generate or access content sexualizing minors will see an alert and be redirected to specialized support resources. Their account is suspended for 24 hours. Repeated violations can result in permanent removal.
Psychologist Joel Scanlan at the University of Tasmania independently evaluated the warning messages in a trial. Organizations including Protect Children and the Lucy Faithfull Foundation contributed to the content of those messages.
Developer Guidelines and Law Enforcement Tools
OpenAI also published updated guidelines for developers building on its platform, covering how to prevent, detect, and report CSAM across API-based products. Additionally, the company is providing law enforcement investigators with AI-powered tools, training materials, and a field guide for conducting online child exploitation investigations.
Context: Growing Pressure on AI Companies
The release comes as OpenAI faces several active lawsuits relating to alleged harm to minors. Common Sense Media has labeled OpenAI's technology an «unacceptable risk» for teens. Rival AI platform Grok has faced its own criticism after being used to generate nonconsensual sexualized imagery, including of children.
The CSAM classifier represents OpenAI's most technically specific child-safety offering to date, moving beyond policy commitments toward deployable infrastructure that other organizations can integrate. Whether the broader AI industry follows with comparable tools — particularly open-weight model providers — remains an open question.
As first reported by TechCrunch, OpenAI has spent much of 2026 strengthening its child safety posture: releasing a Child Safety Blueprint in April, retuning its models, and launching a teen-specific ChatGPT variant. Sunday's announcement marks a shift from internal guardrails toward industry-wide tooling.
Originally reported by TechCrunch. Read the original article for additional details.
View original source