# Spain - VeriPol NLP for false denunciations stopped by the National Police
Spain, 2019 | Severity: N2

## What Happened
In 2019, the Spanish National Police trialed an AI-driven natural language processing (NLP) tool called "VeriPol" to detect false police reports regarding thefts and robberies. The algorithm analyzed the text of police reports to identify specific linguistic patterns associated with fabrication. While police praised its speed, civil rights and legal groups raised serious concerns about high error rates and systemic bias. The algorithm systematically flagged reports submitted by low-income individuals or those with speech differences as "false denunciations" based on linguistic proxies. Due to these concerns and legal challenges regarding procedural fairness, the National Police stopped using the system for automated decision-making.

## How MEG 1 acted
VeriPol failed both the Dynamic Accuracy Index (DAI) and the Index of Safety and Responsibility (ISR) audits due to its high error rate and risk of systemic bias, violating the principles of informational reliability and procedural fairness.

## MEG 2 Analysis - Legal Framework
* **Agent level:** N2 (management) - automated decision-support and text analysis system used by a public authority to make critical law enforcement and credibility decisions.
* **Cause (6.1):**
  * **(a) system defect** - algorithmic bias and flawed linguistic indicators built into the NLP model (design defect), attributable to the developer.
  * **(b) autonomous decision error** - the system's decision to flag legitimate reports as "false denunciations" based on arbitrary proxy variables.
* **Attachment of liability:** The Spanish National Police as the operator is fully liable for the wrongful classifications and investigative delays; the developer is liable for design defects (6.1a).

## Procedural Mechanism
* **6.2 (architectural human confirmation):** Decisions involving the credibility or classification of police reports must require active, architectural human confirmation. The AI can only flag discrepancies; a human officer must verify the recommendation.
* **7.1 (independent forensics):** All analysis parameters, matching scores, and the system's decision path must be securely logged in an independent ledger.
* **9.4 (least privilege):** The scoring system was granted significant technical authority to alter taxpayer/citizen credibility statuses without independent validation, violating safe operational parameters.