The promise of AI in policing is shadowed by a stark admission: the technology itself carries inherent bias. This isn't a hypothetical concern; it's a reality that a leading police chief is now openly acknowledging, while simultaneously vowing to actively combat its risks. The push for greater AI integration in law enforcement across England and Wales, championed by Labour and supported by many police leaders, aims to equip officers with the cutting-edge tools needed to stay ahead of evolving criminal threats. But here's where it gets controversial: can we truly trust AI to be impartial when its very foundation can be built on flawed historical data?
Alex Murray, director of threat leadership at the National Crime Agency and the national lead for AI, has been candid about these challenges. He revealed that a new national police AI center is being established with the explicit goal of recognizing and minimizing these inherent biases. Think about it: when algorithms are trained on past data, and that data reflects past societal prejudices, the AI can inadvertently perpetuate those same unfair outcomes. This could mean certain communities being disproportionately targeted by surveillance or individuals being misidentified based on their race, gender, or economic background. And this is the part most people miss: it's not just about identifying bias; it's about how we train our officers to interpret and act upon the AI's outputs to ensure fairness.
Murray emphasized that technologies like live facial recognition and predictive policing, while powerful, are indeed prone to bias. He stressed the crucial role of data scientists and engineers in meticulously cleaning the data, training models correctly, and rigorously testing them. "There is no point releasing something to policing that has bias in it that’s not recognised, and everything should be done to minimise it to a level where it can be understood and mitigated," he stated. This is a critical point for all of us to consider – are we comfortable with technology that might not be fully understood or mitigated?
We've already seen examples of this bias in retrospective facial recognition systems, where a suspect's image is compared to a database after a crime has occurred. Live facial recognition, a more contentious and less frequently used tool, operates in real-time to identify suspects, and it too, is not immune to bias. A report from December highlighted how a retrospective facial recognition system used by police had inadequate safeguards in place. The Association of Police and Crime Commissioners (APCC) pointed out that these system failures were known for some time but weren't adequately communicated to affected communities or key stakeholders. Darryl Preston, the APCC's forensic science lead, underscored the necessity of independent oversight for these potent tools, stating, "It is not acceptable for technology to be used unless and until it has been thoroughly tested to eliminate bias. That clearly was not the case in this instance." Is this a failure of technology, or a failure of oversight?
The new national AI center, backed by a substantial £115 million investment, aims to address these issues by reducing bias and vetting products from private vendors. Currently, individual police forces make their own decisions about AI adoption, a process seen as slow and inefficient. Murray likened the situation to an "arms race" with criminals who are also leveraging AI. He cited a disturbing case where a paedophile attempted to use a deepfake image to evade conviction for child abuse, forcing police to expend resources disproving the fabricated evidence.
Murray also moved to dispel the common, often sensationalized, association of AI in policing with dystopian futures like 'Minority Report.' He argued that AI's benefits extend far beyond predictive policing, offering significant assistance across a wide spectrum of crimes and challenges. However, he reiterated that the ultimate decision-making power must always remain with human officers.
Looking ahead, AI could be instrumental in combating disinformation campaigns that aim to incite public unrest, assist in manhunts, expedite the search for vehicles linked to suspects, and drastically reduce the time detectives spend sifting through vast amounts of CCTV footage or seized digital devices. "What took days, weeks, sometimes months can potentially take hours," he explained.
A recent case involving cashpoint attacks illustrated AI's transformative potential. Police in Luton arrested four suspects and, by using AI to analyze data from their phones, secured guilty pleas within weeks. The AI not only processed data in Romanian but also translated it, identified potential criminal activity, and compiled a comprehensive package for detectives. Trevor Rodenhurst, chief constable of the Bedfordshire force, attested to the efficiency gains, stating, "This allowed us to draw evidence from lots of devices with a vast quantity of data, which we would otherwise not have been able to do." He further noted a shift in frontline officers' perspectives: "They are no longer suspicious, they are asking when they can have it. That capability is transformative."
Considering these advancements, but also the acknowledged biases, where do you stand on the expanded use of AI in policing? Are the potential benefits worth the risks, and what measures do you believe are most crucial for ensuring fairness and accountability? Let us know your thoughts in the comments below!
