On November 16, 2026, OpenAI announced four initiatives building on its existing Child Safety Blueprint and ongoing detection and reporting systems (alongside the OpenAI CSAM Image Classifier Model Card published the same day). Rather than a general-purpose model launch, the update focuses on safety tooling, in-product interventions, developer guidance, and law enforcement support.

Free API Image Classifier with AI-Generated Signal for Vetted Organizations

Classifier Capabilities & Access Scope: OpenAI is offering vetted child protection organizations and industry safety teams free access via the OpenAI API (subject to intake review) to a new image classifier that detects potential CSAM, including AI-generated material, and provides a dedicated signal indicating whether material may be AI-generated. Developed with input from NCMEC, the classifier outputs scores to help human reviewers prioritize queues, and OpenAI has partnered with Tech Coalition and ROOST as well as joined Tech Coalition's Lantern signal-sharing program.

Official & Early Partner Testing Metrics: In early internal tests on user-uploaded images reported by OpenAI, the classifier flagged 48% fewer images while identifying roughly the same volume of abusive material, making flagged images about twice as likely to contain abuse. Early partners also reported their own evaluations: fal confirmed 72% of high-confidence flags on human-reviewed production images were previously unseen CSAM, invideo AI noted roughly one-quarter of flagged cases led to bans, and SafetyKit reported 90% detection on sexualized drawings or animations alongside less than 1% false positives on legal adult content.

ChatGPT Product Interventions, Developer Guidance & Law Enforcement Support

Warnings, Conversation Deletion & 24-Hour Suspensions: Building on existing prohibitions and refusals, OpenAI is building in-product ChatGPT warnings to interrupt searches for abusive material, explain the harm (including when AI-generated material is requested), and direct users to specialist help resources. The system will delete flagged conversations and pause ChatGPT access for 24 hours while preserving access to specialist support; repeat violations may lead to longer restrictions, account termination, or reports to authorities.

Developer Guidance & Authorized Law Enforcement Tools: OpenAI also published developer CSAM guidance (informed by Tech Coalition, Thorn, NCMEC, and IWF, alongside previously released Tech Coalition disclosure guidance co-commissioned with Anthropic and Google) and began offering program-managed investigative tools and training to authorized child protection law enforcement agencies to review long chat logs and legal returns with source-linked timelines.