A WIRED investigation into a British police region's predictive analytics system reveals significant accuracy problems with the AI tool designed to forecast criminal activity.
UK law enforcement agencies have increasingly adopted artificial intelligence for crime prediction, but a detailed examination shows the technology delivered inconsistent and questionable results in at least one major implementation.
The system, deployed across a police region, aimed to identify high-crime areas and patterns before they occurred. However, testing and analysis found the predictions often failed to match actual crime data, raising questions about the reliability of such tools in operational use.
The investigation highlights broader concerns about how AI systems are deployed in policing without adequate validation. Experts note that flawed predictive models can perpetuate bias and misdirect resources away from genuine hotspots.
Police departments continue expanding AI adoption despite these gaps. The findings underscore the need for rigorous testing protocols and transparency around algorithmic decision-making in law enforcement. Officials have not yet detailed plans to address the accuracy issues identified.
A security researcher has developed an algorithm that generates computer-generated patterns capable of evading detection by surveillance cameras. The technique can hide people, faces, and vehicles from AI-powered monitoring systems.
Scammers are enrolling fake students at US community colleges, using artificial intelligence to complete coursework, and collecting financial aid payouts. The scheme exploits gaps in enrollment verification and assignment monitoring.
The Head Mare hacktivist group has compromised TrueConf video conferencing servers and replaced legitimate client installers with trojaned versions containing backdoors.
OpenAI inadvertently launched a denial-of-service attack against Hugging Face, the popular machine learning platform. The incident has prompted questions about AI infrastructure security and unintended consequences of large-scale operations.