Introduction
Our digital unit was small and overstretched, with no overtime and a backlog that never stopped growing. When the warrant came through — six devices, including a NAS — we knew we’d be stretched thin. By the time imaging finished, we were looking at over 600,000 images and 50,000 videos.
This wasn’t a straightforward CSAM case. The content was personal, unstructured, and offered few known hits. We expected weeks of manual sorting and flagged the case as high-risk for delays.
We expected to spend the month just getting through the noise. Instead, we got straight to what mattered.
Rather than diving straight into filters and face grouping, we started by enabling the S21 Global Alliance Database (GAD) inside S21 VisionX and let it run before we touched a single frame.
Within minutes, over 90% of the dataset was matched and categorised. But this wasn’t just binary yes/no labelling, it was contextual tagging across public adult material, legal family footage, known CSAM and previously seen, but non-illegal, content.
Each file came with supporting case history, investigator notes and cross-device linkage. We weren’t just scanning, we were stepping into a map that had already been partially drawn.
That left us with only 10% of files for hands-on review. That’s where we found what mattered.
One clip flagged under offline CSAM detection showed a young child. Not in the S21 GAD, not previously known. Age detection highlighted the concern and further review confirmed it: a first-generation victim, previously unreported.
Because S21 GAD had cleared the bulk of the noise, we had the clarity, capacity and time to find them.