How AI Models Learn Bias From Police Body Camera Footage

AI Ethics with Fexingo: Bias, Safety, and Responsible Artificial Intelligence · Jul 12, 2026 · 8m

In this episode of AI Ethics with Fexingo, Lucas and Luna dive into a growing concern: biased AI analysis of police body camera footage. They examine a 2025 study from the University of Michigan that found speech-to-text models performed 20% worse on officers from underrepresented backgrounds, leading to inaccurate incident reports. The hosts discuss how this bias can compound in evidence analysis, affect public trust, and what steps researchers are taking to improve datasets. A concrete look at how technology meant to increase accountability can inadvertently embed new forms of bias…