VeriPol flags 25 and 39 false robberies in Murcia and Málaga

AI system

System: VeriPol
Model: Not assessed
Release: Not assessed
Application developer: Universidad Complutense de Madrid and Universidad Carlos III de Madrid, with Spanish National Police officer Miguel Camacho-Collados
Role: Not assessed
Configuration: Not assessed
Inputs and outputs: Not assessed
Autonomy: Not assessed
Workflow: Training on 1,122 robbery reports from 2015 (534 genuine, 588 false) labeled by one experienced officer. Text features selected by L1 penalization plus heuristic rules, scoring reports at the police station.
Human oversight: Not assessed
Relation to AI: The peer-reviewed paper describes a natural-language-processing and machine-learning classifier with L1 feature selection that outputs the probability a report is false. The paper's authors include the police developer. AlgorithmWatch and Civio report that no independent verification was done before the police stopped using it in October 2024.
AI contribution: The peer-reviewed paper describes a natural-language-processing and machine-learning classifier with L1 feature selection that outputs the probability a report is false. The paper's authors include the police developer. AlgorithmWatch and Civio report that no independent verification was done before the police stopped using it in October 2024.

Outcome identifiers: Not assessed
Benefit: 25 and 39 false robberies respectively, against 3 and 12 in the same month
Category: Reduced Risk
Status: Not assessed
Beneficiary groups: Organizations. Society.
Beneficiaries: Not assessed
Reach: Not assessed
Distribution and access: Not assessed
Duration: Not assessed
Expected?: Yes
Costs and harms: Not assessed

Studies: Not assessed
Study design: Not assessed
Measurement: Not assessed
Setting and sample: Not assessed
Measures: 25 and 39 false robberies respectively, against 3 and 12 in the same month
Comparison: false robbery reports detected by officers in the same month over the previous decade
Results: Reference: only three and 12, respectively, for that month in the previous decade. Result: 25 and 39 false robberies in just one week.
Additional value from AI: Not assessed
Limitations: Not assessed
Denominators: Not assessed
Materials for replication: Not assessed

Participants: Developer: Universidad Complutense de Madrid and Universidad Carlos III de Madrid, with Spanish National Police officer Miguel Camacho-Collados. Operator: Spanish National Police (Policía Nacional), Ministry of the Interior.
Jurisdictions: Benefit: Murcia and Málaga, Spain. Developer: Spain. Operator: Spain.
Ownership and control: Not assessed
Dependencies: Not assessed
Cross-border connections: Not assessed
Geopolitical implications: Not assessed

Conditions for replication: Reports written in the same language register as the training data (the Valencia report faulted missing regional variation). Officer training and protocols for acting on scores. Independent field evaluation, absent before discontinuation.
Replication attempts: 3 follow-up entries. Outcome unresolved: 2. Benefit did not recur: 1.
Later findings: 25 March 2025. Civio reported that the National Police stopped using VeriPol in October 2024 (day not given) because, the Interior Ministry said, it lacks validity in judicial proceedings.
Availability: Not assessed
Institutional decision: Not assessed
Grounds: Not assessed
Action and result: Not assessed
Success criterion: Not assessed

Assessed benefits: Organizations. Society.
Depth: Not assessed
Reasons: Organizations: Flagged-report counts and confession rates are police aggregates with no independent confirmation. Society: No measured effect on insurance fraud or police resources is documented.
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: Independently reported

Kind: Not assessed
Related entries: Not assessed
Field: Fraud and false reporting. False robbery reports filed with police.
Purpose: Estimate the probability that a robbery report is false from its text to support police decisions.
Dates: Date basis unassessed: 2019 (precision unassessed). Updated: 22 October 2026.
Stage: Not assessed
Version: SHA-256: 03c957adc9b1…