Checkstep has secured access to OpenAI’s new CSAM Image Classifier - announced on 16th November - and can now make it available to customers, expanding the specialist detection options platforms can use to tackle child sexual abuse material.
The classifier provides separate signals for potential CSAM (child sexual abuse material) and potential AI-generated CSAM in still images. These signals help teams prioritize content for review.
Why this matters
Every CSAM image represents the abuse of a child, and every time it is shared, that harm is repeated. Platforms have relied on matching uploads against databases of known material, and that remains essential. But the challenge is changing fast. The Internet Watch Foundation assessed 6,310 AI-generated images meeting the legal definition of child sexual abuse in the first half of 2026 alone, already more than the 4,512 it found across the whole of 2025. IWF experts warn that rapidly improving technology is making this material indistinguishable from genuine imagery. Because each AI-generated image can be brand new, it may not appear in any database of known material. That is why detection now needs layers: matching what is already known, identifying what is new, and connecting both to trained human review and decisive action.
Building on Checkstep’s existing capabilities
Checkstep already integrates with C3P’s Project Arachnid and Resolver’s Athena. Together, these services address different detection needs.
- Project Arachnid’s Shield service compares media against digital fingerprints of known material, helping platforms identify and disrupt its recirculation.
- Athena uses a classifier trained on the UK’s Child Abuse Image Database to flag potential previously unseen abuse material, including synthetic and AI-generated imagery.
- OpenAI’s classifier adds another option for detecting potential material that known-content matching alone may miss. Its value alongside existing tools should be assessed on representative platform traffic.
Connecting detection to action
Effective child-safety operations depend on what happens after a flag. Checkstep connects detection with configurable policies, human review, enforcement, escalation, and reporting tools, helping teams manage their response and monitor its quality.
A classifier flag does not confirm abuse or determine reporting obligations. Trained review and established response processes remain essential.
Read about our existing capabilities here.
Read about the classifier here.
If CSAM detection is a priority within your roadmap, this is something you will want to know more about. To learn more about how OpenAI’s classifier can add to your platform, book a demo, and we’ll walk you through evaluating it alongside your current detection.