OpenAI releases free CSAM image classifier for vetted organizations
Original title: Strengthening the fight against online child sexual abuse with new tools and support
Early internal tests estimated 48% fewer flags while finding about the same number of abusive images.
OpenAI said on November 16 it is offering a new image classifier through its API at no cost to vetted organizations, with assistance from NCMEC. The tool looks for visual signs of abuse, including AI-generated material, and returns scores so reviewers can prioritize images. OpenAI also described interventions, developer resources, and support for investigators, and said it joined the Tech Coalition’s Lantern program. The Tech Coalition and ROOST will help share access.
Unconfirmed: Detection figures are OpenAI’s early estimates; partner evaluation results are cut off in the source.
- OpenAI calls it the first broadly available CSAM classifier with an AI-generated signal.
- In one evaluation, flagged images were about twice as likely to contain abuse.
- NCMEC assisted with training, evaluation, and real-world deployment design.
- OpenAI supports Thorn’s Safer and began participating in Lantern.
- Vendors
- OpenAI
- Organizations
- NCMECTech CoalitionROOSTThorn
From the original article
11 min read at openai.comNew tools, practical guidance, and partnerships will help prevent, detect and respond to online child sexual exploitation and abuse.
openai.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.