French Tax Authorities Harness Artificial Intelligence to Combat Fraud
The French tax authorities are continually refining their artificial intelligence (AI) technology to improve its ability to detect potential discrepancies and combat fraud.
A Significant Increase in AI-Initiated Investigations
The ‘ciblage des opérations fiscales’ service, part of the Direction générale des finances publiques (DGFIP), has seen a significant increase in AI-initiated investigations into possible business fraud. Up to half of all investigations now originate from the system.
How the AI-Powered System Works
The AI-powered system uses algorithms, tax identification numbers, confidential tax data, and mathematical formulae to highlight potential fraudulent activity. Managed by data scientists, the system is housed in a secure location in France to prevent attacks. The AI algorithms are designed by 32 specialists, including contract data scientists, state (INSEE) researchers, and tax specialists.
Identifying Anomalies and Launching Tax Audits
The system works by scanning thousands of pieces of data through ‘data-mining’ to identify files that present an ‘anomaly’ and therefore have a greater chance of being fraudulent. Identified cases are then sent to the local finance authority, which launches a tax audit.
- 50% of all business checks are now made as a result of AI
- The remaining 50% come from more traditional detection methods such as tax agents’ suspicions or tip-offs
Targeting Fraudulent Cases Among Individuals
The DGFIP is also aiming to target possible fraud cases among individuals by 2027, focusing on ‘high-stakes cases’ such as those with high assets and property wealth. The technology cross-references all available data to predict how much rent should be declared, flagging any significant discrepancies.
Additional Successes of the AI System
AI has already been used to help with other financial documents, including:
- Allocating amendments to finance bills, which previously required 10 agents working for 24 hours but was accomplished by the algorithm in just two hours
- Policing other forms of declarations in France, identifying nearly 120,000 potential cases of undeclared swimming pools and €40 million to €50 million in additional tax revenue
Challenges in Extending the System
However, extending the system to identify undeclared buildings is more challenging due to the complexity of the task. The authorities are currently experimenting with using aerial photos to check for unregistered structures, including garden sheds and extensions, but the process is slower and less accurate than having surveyors in the field.
Note: This article has been formatted according to Markdown syntax, which uses headings (#), subheadings (##), bullet points (*), and plain text.