待翻译：England and Wales: Growing gap between power and regulation of police AI

AI 服务暂时不可用，以下为来源正文，待恢复后补全翻译。

Back to articles News England and Wales: Growing gap between power and regulation of police AI Police and legal agencies in England and Wales are actively using at least 27 AI technologies, according to an academic study. Date 5 August 2026 Author Chris Jones Topics Snooping and surveillance Region United Kingdom (UK) Police and legal agencies in England and Wales are actively using at least 27 AI technologies, according to an academic study. An academic project has found that at least 27 artificial intelligence (AI) tools are in active use by English and Welsh police forces and agencies dealing with prosecution, imprisonment and probation.A further 27 have been tested in trials and more than a dozen others have been the subject of unspecified “experiments” or used in “unknown” ways.. The 70 uses of AI in the criminal justice system mapped by the Probable Futures project range from emergency call handling and evidence assessment, to crime ‘prediction’ and automated analysis of video footage. Of the tools documented, 41 can be used for analysis, 19 are capable of synthesis (“combines inputs to enrich or predict”) and 28 can be used for generation (“creates new content from prompts or inputs”). A single tool may have more than one capability. A report published by the project in June warns of “an increasing gap between technological capability and the conditions required for its legitimate, reliable, and responsible use.” Police acquire technology without “understanding the capabilities” There is no specific legal basis governing police use of AI technologies in England and Wales.1 Police forces are acquiring powerful new technologies “without fully understanding the capabilities… and the implications,” according to William Webster, the government official responsible for oversight of biometric and surveillance technologies. Despite the existence of a “national decision making model” for buying new technology, Webster said his office has found “no evidence” of policing forces following it. The government is aiming to introduce centralised technology procurement for the police, which could improve transparency and public oversight. Plans for controlling the actual use of those technologies are more limited. Technology outpacing the law Later this year, the government is due to introduce legislation to govern the police use of facial recognition technology. However, the data collected by Probable Futures shows facial recognition is merely the tip of a much larger iceberg that already includes tools used for extremely sensitive aspects of police work. Dr Temitope Lawal, a Research Fellow at Northumbria University and member of the Probable Futures project, told Statewatch that some AI tools are “delivering genuine benefits.” However, he also warned that tools for “shaping evidence and/or case progression are being introduced through fragmented approaches, while the legal and governance frameworks have not developed at the same pace.” Police AI tools Lincolnshire Police are using a tool called CESIUM, a “cloud-based IT solution that utilises Artificial Intelligence and Machine Learning to run over force systems to readily identify and flag risk of Child Exploitation.” Avon and Somerset Police use Qlik Sense, which the force says is “primarily used as a visualisation tool to monitor demand, trends and harm.” Probable Futures say it is used for risk prediction – that is, to calculate the risk someone will commit a criminal offence or someone will become a victim of an offence. There appears to be significant enthusiasm for ‘predictive’ technology at Avon and Somerset Police, as there is at other forces in the UK and across Europe. A recent investigation published by Wired found many of the tools tested and deployed by Avon and Somerset have been riddled with errors. West Yorkshire Police are also using ‘predictive’ technology, in the form of a tool called Corvus IOM Case. This produces “an individualised score aimed at providing an indication of an individual’s likelihood to reoffend,” according to Probable Futures. Police are also using systems to ‘predict’ burglaries and the likelihood of “investigative success for minor, non-domestic assault and public order offences,” amongst many others. The risks of generative AI Dr Evdoxia Taka, a Research Associate at the University of Glasgow who works with the Probable Futures project, emphasised the police’s growing use of generative AI systems.2 Of the dozens of tools mapped, 40% use generative AI “to perform at least one task, such as document summarisation,” Taka told Statewatch. These systems could “inadvertently add content to a witness statement, causing an otherwise weak case to appear stronger,” she warned. “This could lead to a charging decision where the appropriate outcome would otherwise have been no further action.” Generative AI tools can also be used for deliberate falsification. An officer with Derbyshire police has been accused of using generative AI to “create evidential material in a number of cases,” and faces accusations of perverting the course of justice. A month after that case made headlines, the government announced plans for all police forces to use AI “to review and summarise evidence” by 2027. Surveillance technology made in Israel Alongside concerns over the chain of evidence in criminal investigations are problems with digital technology supply chains. These require destructive mining for raw materials, huge amounts of energy and water, and often have their origins in war and oppression. West Midlands Police use video analytics software produced by a company called Briefcam to “make recorded footage searchable and help investigators identify relevant information,” according to Probable Futures. Cumbria Police are also known to have a contract with the company. Briefcam is an Israeli-founded company that was later acquired by Milestone Systems, a Danish firm. Al Jazeera note that the company’s technology “has been used in East Jerusalem, Palestinian territory illegally occupied by Israel.” Israeli technology is also powering the growing use of facial recognition cameras by police forces in England and Wales. A company called Corsight AI was subcontracted to supply software that will soon be used by every police force in the two countries. An incomplete picture Policing in England and Wales is highly decentralised, with dozens of regional police forces working alongside a variety of national bodies. It is highly likely there are other AI tools in use that have not been documented by Probable Futures. In a white paper published earlier this year, the government said it would create a national register of AI tools used by the police. A recent government “factsheet” suggests this register will not be comprehensive. The register is “expected to align with FOI [freedom of information] exemptions,” says the document. This means that information could be excluded in the name of protecting law enforcement investigations, commercial interests or national security. It also appears that completing the register will not be compulsory. “The government expects all Chief Constables to complete the registry once it becomes available and will act where this is not the case,” says the document. Wildfire and heroin The head of the College of Policing, Andy Marsh, has said that the new National Centre for AI in Policing (‘Police AI)’ must test tools “with rigorous evaluation, and then spread them like wildfire through policing.” He has also said the police need AI injected into their daily work “like heroin into the bloodstream.” It remains to be seen whether Marsh’s efforts will be as uncontrolled, destructive and addictive as either wildfire or heroin. Image credits: “Policing Minister Sarah Jones speaks at Police AI launch event” by UK Home Office, CC BY 4.0 England and Wales have a separate criminal justice system to Scotland and Northern Ireland. Generative AI systems are those that create textual, audio or visual material, such as ChatGPT.