Practitioners see patterns of harm every day. A teacher notices manipulative AI targeting students. A parent discovers unconsented data collection. A community organizer watches misinformation spread through networks platforms don’t monitor.

These observations are fragmented. They stay local. They don’t aggregate into the kind of evidence that decision-makers are required to act on. Meanwhile, companies conduct risk assessments using internal data, academic proxies, and filtered advocacy. The knowledge that would make those assessments credible sits with practitioners, but there’s no mechanism for that knowledge to travel.

This talk presents SIGNAL: a system that allows practitioners to report observations in standardized formats that can be aggregated across communities and contexts. I walk through why current approaches fail, what changes when ground-level intelligence becomes structured evidence, and what it takes to build infrastructure that shifts who gets to define what counts as harm.