AI enabled policing platforms (FOI ref: 1948A/25)

Request

I am writing to request information under the FOIA which relates to predictive analytics, data intelligence or AI-enabled profiling that have been in trial, development, or deployment at any point in 2025.

1. Copies of any portfolio, project listing, reports, presentations, or internal summaries relating to the above that are currently, or have been, in development, trial, or operation within this police force, or its associated ROCU.

2. Where available, details of:

The name or description of each project

The units or departments involved

External collaborators (contractors, researchers, consultants, etc.)

The intended purpose or application of the technology

The status (e.g. research, pilot, deployment, discontinued)

Any funding source (Home Office, NPCC, or other)

Response

Question 1 – Under section 21 of the Act, we are not required to provide information in response to a request if it is already reasonably accessible to you. The information you requested is available on the ‘Ethics Panel Reports and Minutes’ section of the West Midlands Police and Crime Commissioner’s website: https://www.westmidlands-pcc.gov.uk/governance/ethics-panel/

For example: Ethics Committee – February 2023: Agenda Item 4 - Theft of Motor Vehicle (final report)

Question 2 - There are currently four machine learning models that are being used to provide information by way of making predictions.

These are:

1. Short-term knife crime (used causing injury affecting young people) which provides a category of relative likelihood of a knife crime occurring by geography (a 1 square kilometre grid) over a coming 4 weekly period. This is used to inform deployment / patrolling and resourcing decisions by the Guardian project team.

2. Theft of motor vehicle; this provides a category of the relative probability of a vehicle theft occurring over the coming month by geography (0.75 of a square kilometre) covering 8 different vehicle types. This has been used to inform deployment decisions.

3. There is also the Seasonality planner which provides an indication of the expected relative occurrence for each month of the coming year (is a crime type expected to improve or get worse) and provides forecasts for 12 months for each local policing area and the whole force area. This covers 26 crime types.

4. Stalking and Harassment and harm escalation prediction; this relates to people who have undertaken stalking and / or harassment and estimates the probability that they will go on to commit high harm crimes. A similar model is also run relating to people who have been victims of stalking and harassment (it estimates the probability that they will be victims of high harm crimes).

These have been created by the in-house Data Analytics Lab.