# What if AI can help us fight organized crime and corruption? | Gian Maria Campedelli | TEDxGeneva

Source: https://www.youtube.com/watch?v=Dh6TV1RQ_so
Recap page: https://rapidrecap.app/video/Dh6TV1RQ_so
Generated: 2025-12-19T17:44:02.436+00:00

---
## Quick Overview

Artificial Intelligence (AI) and machine learning tools offer a proactive, data-driven approach to fighting organized crime and corruption, especially in contexts like Italy where mafias have historically infiltrated local governments, by analyzing public data to predict infiltration risk two years in advance, which is more effective than reactive state responses.

**Key Points:**
- Italy has a high incidence of mafia infiltration, resulting in 368 city councils being dismissed between 1991 and 2025, affecting over 3 million people.
- The traditional state response is reactive, allowing corruption to become entrenched; AI offers a proactive solution to predict infiltration.
- The proposed AI approach relies on open-access data, specifically analyzing how public money is spent (e.g., on tourism vs. politics) and political characteristics of mayors/elections.
- The initial model successfully predicted infiltration two years in advance with 75% accuracy, significantly better than the state's reactive methods.
- The AI tool focuses on the top 5% riskiest municipalities to detect 90% of infiltrated municipalities, maximizing resource efficiency.
- The goal is not to replace humans but to empower citizens and law enforcement by making data central to decision-making and increasing transparency and privacy safeguards.

![Screenshot at 00:04: The opening slide displays the theme of the TEDxGeneva event, "CHAOS," scheduled for November 28, 2025, setting the stage for a discussion on complex, systemic issues like organized crime.](https://ss.rapidrecap.app/screens/Dh6TV1RQ_so/00-00-04.png)

**Context:** Gian Maria Campedelli presents a TEDxGeneva talk on leveraging computational criminology and Artificial Intelligence (AI) to proactively combat organized crime and corruption, focusing on the specific challenge of mafia infiltration into Italian municipal governments following a period of intense confrontation in the early 1990s. He contrasts the slow, reactive measures taken by the state with the potential of using machine learning on publicly available data to predict where corruption is likely to occur, thereby shifting the approach from reaction to prevention.

## Detailed Analysis

Gian Maria Campedelli argues that AI can significantly aid in fighting organized crime and corruption by enabling proactive measures rather than reactive ones. He highlights the severity of the problem in Italy, citing that between 1991 and 2025, 368 city councils were dismissed due to mafia infiltration, affecting over 3 million people. He notes that the state's response is inherently reactive, often intervening only after damage is done. Campedelli, a computational criminologist, proposes using machine learning—specifically open-source models—to analyze public data (like public spending allocations and political characteristics) to predict which municipalities are at the highest risk of infiltration. The initial model achieved a 75% accuracy rate in predicting infiltration two years in advance, allowing law enforcement to intervene proactively. This method prioritizes resources by focusing on the top 5% of high-risk municipalities, where 90% of infiltration is detected. The core philosophy is using AI to make democracies stronger and fairer by increasing transparency, protecting privacy, and ensuring that decision-making tools are centered on citizens, not just law enforcement needs.

### The Problem

- Mafia Infiltration in Italy: 368 city councils dismissed for mafia infiltration between 1991-2025
- 3+ million people affected
- State response is reactive, not proactive
- Mafias have changed behavior, embedding themselves in politics and the legal economy.

### The Proposed Solution

- AI and Open Data: Use computational techniques on open-access data (public spending, political characteristics) to predict infiltration risk two years in advance
- The core idea is f(public_spending, political_characteristics) = corruption_risk.

### Key Results and Efficiency

- Initial model achieved 75% accuracy in predicting infiltration two years out
- Focus on the top 5% of at-risk municipalities to detect 90% of infiltration, given finite resources.

### The Vision

- A Proactive Approach: Move from reactive policy to proactive prevention
- Empower citizens and make democracies stronger and fairer by ensuring data-driven tools protect privacy and keep citizens central.

![Screenshot at 00:00: Introduction slide showing the TEDxGeneva branding.](https://ss.rapidrecap.app/screens/Dh6TV1RQ_so/00-00-00.png)
![Screenshot at 00:09: Speaker Gian Maria Campedelli introduces the core question: "What if AI can help us fight organized crime and corruption?"](https://ss.rapidrecap.app/screens/Dh6TV1RQ_so/00-00-09.png)
![Screenshot at 00:15: Slide displaying the flags of Italy, Switzerland, and the UK, prompting the audience to guess which has the lowest homicide rate.](https://ss.rapidrecap.app/screens/Dh6TV1RQ_so/00-00-15.png)
![Screenshot at 00:49: Slide showing homicide rates comparison \(Italy 0.57, Switzerland 0.60, UK 1.11, USA 5.76, Jamaica 49.30\) to illustrate the contrast in violence.](https://ss.rapidrecap.app/screens/Dh6TV1RQ_so/00-00-49.png)
![Screenshot at 03:10: Slide summarizing the problem: "1991-2025: 368 city councils dismissed for mafia infiltration. 3+ million people affected."](https://ss.rapidrecap.app/screens/Dh6TV1RQ_so/00-03-10.png)