evidoria

← Back to browse

Good practice Imported

The Homicide Prediction Project — the UK Ministry of Justice's Secret AI Risk-Profiling Tool

United Kingdom · London · See the United Kingdom profile · See the London profile

Evidence: Descriptive / self-reported Top 99% 13/100 · Ask Evidence Copilot about this practice

FOI documents revealed the UK Ministry of Justice built a predictive-policing tool with the Home Office and two police forces, using data on 100,000-500,000 people, including mental-health and self-harm records, to flag those 'at risk' of committing homicide.

100,000-500,000
People whose records were shared to build the tool (2023-2025)
1,300+/day
OASys (separate MoJ tool) daily risk assessments
The Homicide Prediction Project — the UK Ministry of Justice's Secret AI Risk-Profiling Tool

Details

Maturity
Pilot
Promoter
UK Ministry of Justice, with the Home Office, Greater Manchester Police and the Metropolitan Police
Period
2023-2025
Keywords
justice, criminal justice, policing, data protection, public safety

Context

Documents obtained by Statewatch through freedom-of-information requests revealed that the UK Ministry of Justice, together with the Home Office, Greater Manchester Police and the Metropolitan Police, developed the 'Homicide Prediction Project' (later renamed 'sharing data to improve risk assessment'), intended to identify people considered at risk of committing homicide from government and police records.

Objectives

The stated aim was to build a predictive tool from shared government and police data, drawing on sensitive indicators such as mental-health history, addiction, self-harm and suicide risk, vulnerability and disability, to flag individuals as being at risk of committing homicide.

Activities

A Data Sharing Agreement with Greater Manchester Police shows that records on between 100,000 and 500,000 people were shared to build the tool, and Statewatch obtained the Data Protection Impact Assessment and an internal risk assessment via freedom-of-information requests after the project had not been proactively disclosed.

Results

No independent evaluation of the tool's predictive accuracy or false-positive rate has been published; following press coverage in April 2025, the Ministry of Justice said the project was research rather than an operational tool.

Conclusions

Campaigners and legal experts criticised the lack of prior disclosure, the use of protected health characteristics as predictive inputs, and the absence of published independent evaluation - concerns that echo separate, published findings that the ministry's long-standing OASys reoffending-risk tool, used in over 1,300 assessments a day, shows lower predictive validity for Black, Asian and Mixed-ethnicity offenders.

Implementation

Indicative cost
Medium (€50k–€500k) — No public cost figures disclosed; the project involved data-sharing infrastructure across four government/police bodies.
Time to results
Medium (1–3 years) — Reported period of development 2023-2025; became public through FOI disclosures in April 2025, after which the Ministry of Justice said the project was research rather than operational.
Staffing & skills
UK Ministry of Justice, Home Office, Greater Manchester Police, Metropolitan Police

Conditions for success

  • Formal Data Sharing Agreements between the participating police forces and government departments
  • A Data Protection Impact Assessment and internal risk assessment conducted prior to data sharing

Common failure modes

  • Developing a predictive tool using sensitive health and vulnerability data without proactive public disclosure undermines public trust and legal accountability
  • Relying on protected characteristics (mental-health history, disability) as predictive inputs raises discrimination and fairness risks
  • Publishing no independent accuracy or false-positive-rate evaluation leaves the tool's real-world reliability unverified

Where it fits

Governance type
central government + police, multi-agency
Scale
two police force areas (Greater Manchester, Metropolitan)
Income level
high-income

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful