Researchers warn AI-driven policing could amplify bias against minority communities

Academic research warns that artificial-intelligence systems employed by police risk deepening racial and socioeconomic discrimination. The core issue is that these models learn from past crime data that already reflects human bias, leading to a self-reinforcing cycle where heavily policed areas appear more dangerous. Studies from the University of Sydney Law School, the University of Texas at Austin and New York University document how biased records—sometimes termed "dirty data"—skew predictions and result in disproportionate scrutiny of black neighbourhoods.

While some technologists claim newer AI can adjust for historical bias, they also stress the need for robust human control and transparent oversight. Critics suggest that merely cleaning data is insufficient; independent audits of how the data were generated are required. Without such safeguards, AI-based policing could perpetuate outdated maps of crime, harming already marginalized communities.