WIRED trusts Eticas’ expertise in UK investigation
June 23rd, 2026
WIRED trusted Eticas as an expert to support their rigorous investigative work, and the work has finally been made public!
Avon and Somerset Police built at least 23 predictive models. Some scored close to half a million people. For years, almost no one outside the force knew they existed.
A major WIRED and Liberty Investigates investigation — supported by the Bristol Cable and Lighthouse Reports — has documented what happened next. Eticas reviewed the performance data the force disclosed and found models operating at precision rates below 10 percent, metrics shifting in ways inconsistent with well-governed systems, and bias testing that measured average risk scores by ethnicity without testing for discriminatory outcomes. Two child exploitation models were scrapped. Their source code could not be found.
What the Bristol case makes visible is the distance between deploying an AI system and understanding what it is actually doing. The data existed. The performance records existed. The gap was in who was looking at them, with what methodology, and with enough independence to say plainly what they showed.
High-stakes AI does not fail loudly. It accumulates decisions. The people affected rarely know they were scored.