In brief: A British predictive policing model comprising at least 23 AI models showed no reliable results, highlighting the practical and regulatory limits of predictive policing.
A British AI-powered crime prediction project has failed: at least 23 developed models were unable to deliver reliable forecasts. For a Chief Data Officer, this is a textbook example of the practical limits of predictive policing and the regulatory requirements that such systems must meet.
The British predictive policing project aimed to use machine learning models to forecast crime and thereby deploy police operations more strategically. At least 23 different forecast models were developed and tested, but none delivered reliable results in practice.
From a Chief Data Officer’s perspective, this failure has several regulatory and ethical implications: the use of unreliable AI models in policing carries significant discrimination risks, could lead to unjustified operations in certain districts or population groups, and violates data protection and bias requirements such as GDPR or future EU AI Act provisions. The project demonstrates that pure technical complexity is insufficient when predictive power is lacking.
The failure underscores the necessity of critical governance in AI projects: before models are deployed in safety-critical domains such as policing, strict validation standards, bias testing, and audit structures must be in place. Organisations must recognise early whether a project fails to achieve sufficient hit rates in order to limit reputation and compliance risks.
Source: www.golem.de · Published 25 June 2026
Lumi AI News — AI-assisted curation pursuant to Art. 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.7.1.