evidoria

← Back to browse

Good practice Imported

West Yorkshire Police's AI Domestic-Abuse Risk Tool Tested Inside the ICO's Regulatory Sandbox

United Kingdom · Leeds · See the United Kingdom profile · See the Leeds profile

Evidence: Descriptive / self-reported Top 51% 60/100 · Ask Evidence Copilot about this practice

West Yorkshire Police tested an AI domestic-abuse risk tool inside the ICO's Regulatory Sandbox. The published Final Report shows the biggest gap was transparency planning, not the machine-learning model itself.

Details

Maturity
Pilot
Promoter
West Yorkshire Police, with the UK Information Commissioner's Office (ICO)
Period
2024
Keywords
policing, AI regulatory sandbox, domestic abuse risk assessment, data protection

Context

In late January 2024, West Yorkshire Police (WYP) was accepted into the UK Information Commissioner's Office (ICO) Regulatory Sandbox to test a Domestic Abuse Assessment Tool (DAAT), a supervised machine-learning system intended to help officers assess domestic-abuse risk and flag cases for early intervention, described as 'exceptional innovation using AI in a policing context.'

Activities

Within the sandbox, WYP worked with the ICO on the data-protection questions raised by processing sensitive personal data in a high-stakes policing context; the ICO advised that before WYP could decide what information about the tool's development, accuracy and fairness should be made public, it first needed to complete a stakeholder-mapping exercise.

Results

The engagement prioritized process over immediate outcomes: the ICO's Final Report describes the primary deliverable as 'a clearer transparency roadmap rather than a proven crime-reduction result,' and the tool remained undeployed force-wide at the time of reporting.

Implementation

Indicative cost
Low (< €50k)
Time to results
Short (< 1 year)
Staffing & skills
West Yorkshire Police as promoter, UK Information Commissioner's Office (ICO) Regulatory Sandbox team

Conditions for success

  • Completing a stakeholder-mapping exercise before making public-disclosure decisions
  • Close engagement with the data-protection regulator on fairness/accuracy questions before deployment

Common failure modes

  • Tool remained undeployed force-wide at reporting
  • Sandbox produced a transparency roadmap rather than a proven crime-reduction outcome

Where it fits

Governance type
local/regional police force working with a national data regulator
Scale
regional (West Yorkshire)
Income level
high-income

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.

Similar practices you may find useful