According to the Home Office announcement, PoliceAI will operate as a national centre for the development, piloting and scaling of artificial intelligence across policing in England and Wales. The scheme carries £75 million over three years and is presented as a way to cut administrative time, speed investigations and standardise how forces test new systems. The policy aim is not limited to buying software. Ministers are attempting to create a central route for assessing which tools are accurate, lawful and operationally useful before wider deployment across forces that have historically adopted technology at different speeds.
The £75 million allocation sits inside a wider £140 million AI package for policing over three years. The Home Office said that programme also funds 40 additional live facial recognition units, which would triple current capacity, and a separate £16.5 million package to update how forces handle contact with the public. That second stream covers transcription of 999 and 101 calls, systems that connect crime reports to identify patterns in demand, and tools intended to route non-emergency calls to the right responder more quickly. Read together, the announcement is as much about process change as it is about frontline investigation.
The government’s case for rapid rollout rests on a set of early operational examples. It said AI tools had reduced review of 800 hours of video in a kidnapping case to three hours, which contributed to an early guilty plea, and translated half a million e-books of material in a serious organised crime investigation, leading to arrests. Those examples matter because they target one of policing’s most persistent pressures: the volume of digital material now attached to even routine cases. Mobile phones, messaging platforms, body-worn video and CCTV have expanded disclosure workloads well beyond what many forces can process quickly with existing staffing.
In its first year, PoliceAI is due to focus on areas where the Home Office believes measurable gains can be delivered quickly. Large-scale pilots are planned in up to 10 forces to help officers triage, disclose and summarise digital evidence, with trials running through 2026-27 before a national rollout in 2027 if the results hold. The centre will also build on existing work to redact audio and video files. Ministers said that redaction technology alone could release around 1 million hours each year if adopted by all 43 forces in England and Wales. For chief constables, the practical test will be whether those time savings appear in live casework rather than remaining estimates from pilot conditions.
PoliceAI has also been given a national role in responding to AI-enabled crime. The announcement says the centre will run a Policing AI Threat Hub, with a first focus on deepfake intimate images, and will distribute detection tools and training so forces can identify offences that are becoming harder to prove with conventional methods. A smaller £1 million package is aimed at retail crime and tool theft. The intention is to link police data with property-marking schemes, use AI to identify stolen goods and track resale online, and improve the matching of recovered tools to their owners. If data standards can be aligned, that could shorten the time between seizure and return of property.
The governance offer is central to the announcement. PoliceAI is expected to become part of the planned National Policing Service and, in partnership with CENTRIC at Sheffield Hallam University, it is due to publish a public registry of AI tools used across policing. The first version is scheduled for the autumn. The Home Office said AI models will be independently tested for accuracy and bias, building on work already funded for live facial recognition algorithms. That matters most in functions such as evidence translation and automated summarisation, where any material error could affect disclosure, charging decisions or the reliability of evidence in court. The government has presented those checks as the basis for public confidence in wider deployment.
The College of Policing will host the centre, which places professional standards alongside procurement and testing rather than treating them as separate tasks. Sir Andy Marsh said the College’s role would be to explain how the systems work, how they are evaluated and what safeguards apply, with the Code of Ethics remaining the reference point for operational use. That institutional design is significant. A national centre can reduce duplication between forces, but it also creates a clearer line of accountability for explaining why a tool has been approved, what limits have been set and how officers should record its use.
The launch also sits inside a wider reform programme. The government linked PoliceAI directly to the Police Reform White Paper published in January 2026, the proposed National Policing Service and its wider Safer Streets mission. Sarah Jones, the policing minister, said the programme could free the equivalent of 3,000 officers for visible policing, while the government separately said 3,000 more neighbourhood officers are already in post and 13,000 are due by the end of the Parliament. For forces and police and crime commissioners, the immediate timetable is clear: pilots in 2026-27, a first public registry by the autumn, and a decision on wider scale-up in 2027. For the public, the real test will be less the size of the funding settlement than whether faster evidence handling, clearer oversight and better reporting of accuracy and bias can be shown in day-to-day policing.