Documentary underwriting fraud: how insurers can protect themselves

65 billion euros per year. This is the cost of documentary fraud in France, i.e. 2.5% of GDP¹. The insurance sector alone is losing 2.5 billion euros annually².

11% of French people admit to having already used a fake document³. Among those under 35, this figure rises to 20%⁴.

Fraud has industrialized with generative AI: +244% of digital falsifications Between 2023 and 2024⁵. Environ 50% of insurance scams involve forged documents⁶ : false pay slips, false RIB, false certificates.

Traditional controls are no longer enough. Digital subscription has removed the face-to-face barrier, and falsification tools are available to everyone.

‍

The 3 types of documentary fraud

‍

1 - Falsified documents

The fraudster changes a real document: amounts, dates, beneficiaries. The alterations are imperceptible to the naked eye but detectable by automatic analysis.

Examples: modified bonus/malus certificate (savings 800 €/year), inflated pay slip (2,000 → 4,000€), bank statement with fictitious transfers.

Signals: pixel inconsistencies, modified metadata, outlier fonts.

‍

2 - False documents created from scratch

Integral creation imitating the original structure. Facilitated by generative AI and kits sold for €50-200 on Telegram.

Examples: false proof of address, false garage invoice, false medical certificate.

Signals: invalid SIRET numbers, fictitious coordinates, lack of history.

‍

3 - Synthetic identities

A mixture of real (stolen during leaks) and fictional information. No direct victims, so late detection.

Example: real name + false date of birth + real address + false social security number → subscription life insurance 50K €, premiums paid 6 months, then disappearance.

Why it's dreadful: passes basic KYC checks, no customers complain, organized networks create 50+ identities simultaneously.

‍

Underwriting fraud vs claims

‍

Subscription fraud

The aim is to pay less by reducing the premium. Common methods include the false declaration of the main driver, the omission of a history, the under-declaration of the value of the goods or even a false address in a less risky area. 10% of French people have already declared false information to reduce their premiums.⁷.

‍

Claims fraud

The objective is to receive undue compensation. Fraudsters invent claims, exaggerate damage, or stage accidents. In 2022, 587 million euros of fraud were detected⁸, but the real estimate rises to 2.5 billion euros.

‍

Consequence: premiums are increasing for everyone. Honest customers pay +105€/household/year on average to make up for losses⁹.

‍

Penalties: up to 5 years in prison

‍

On the criminal level, insurance fraud (fraud) is punishable by 5 years in prison and 375,000 euros in fine. The production or use of false documents is subject to 3 years of imprisonment and a fine of €45,000¹⁰.

‍

On the civil level, the consequences are serious: nullity of the contract with retroactive effect, reimbursement of all indemnities paid, termination of all contracts with the insurer, and deletion of the databases. Result: difficulties in finding a new insurer, or premiums multiplied by three¹¹.

‍

Concrete example: An insured person declares his father as the main driver while his 22-year-old son is driving. Six months later, a serious accident occurred. The insurer discovers the fraud and refuses compensation. The customer must reimburse €80,000 in costs incurred for third parties, and faces criminal proceedings: 18 months suspended imprisonment and €5,000 fine.

‍

Why is it exploding

‍

• Digitalization: disappearance of face-to-face, uploaded documents easy to falsify, 100% online experience
• Generative AI: free tools creating ultra-realistic fakes in minutes, deepfakes for biometrics
• Ready-to-use kits: sold for €50-200 on Telegram with French insurance templates, tutorials, support
• Low perceived risk: 73% success rate¹², few convictions, perception “it's not that bad”

‍

Result: +12% /year of fraud¹³. Insurers underestimate the extent: 50% of cases involve documents but little systematic tracking.

‍

5 technological levers

‍

1 - Intelligent document analysis

Automatic detection of manipulations invisible to the naked eye: pixel inconsistencies, modified metadata, aberrant fonts, artificial shadows.

Efficiency: detection 80% + vs 20% manual control, reduction of time 70%, reduction of cost 50%¹⁴

‍

2 - Open Banking

Direct access to bank data (DSP2) to validate real income. A direct connection to the bank cannot be falsified. Use cases: borrower insurance (salary transfers), premium car insurance, life insurance (source of funds).

Advantage: real-time data, 12-24 month history searchable.

‍

3 - Multifactorial biometrics

Video selfie + ID + real presence detection. The customer films for a few seconds while moving his head, the AI compares the face and verifies that it is not a photo.

Efficiency: division by 5 in the fraud rate compared to a simple photo, blocking of 95% + deepfakes.

‍

4 - Data cross-checking

Consistency check via public standards: SIREN/SIRET, La Poste addresses, land registry, driving license.

Example: a customer declares a residential address, but the land register shows an empty lot. Immediate red flag. Minimal cost (Free APIs or less than €1 per request).

‍

5 - Detection of organized networks

Score from 0 to 100 for each subscription based on data consistency, suspicious patterns and digital behavior. Network analysis detects organized groups that increase fraudulent subscriptions.

‍

Example: 15 subscriptions with 15 different names but the same IP address and the same document metadata → automatic preventive blocking.

‍

Checklist protection 2025

‍

Prevention

‍

• Automatic analysis of documents
• Open Banking for income validation
• Biometrics for amounts >10K€
• Data cross-checking
• Team training

‍

Detection

‍

• Automatic score for each file
• Real-time alerts
• Detecting organized networks
• Ongoing account monitoring

‍

Investigation

‍

• Team dedicated to fraud
• In-depth metadata analysis
• Partnership with ALFA
• Liaison with the authorities

‍

Sanction

‍

• Cancellation of the contract if fraud
• Legal action if necessary
• Preventive communication
• Reporting to the ACPR (Supervisory Authority)

‍

3 immediate actions

‍

1 - Audit the real fraud rate

Analyze records from the last 12-24 months with automated analysis tools to quantify avoidable losses and calculate the return on investment.

‍

2 - Deploy automatic document analysis

Start with critical documents: pay slips, bonus/penalty certificates, proof of address. Fast ROI: 50-100K euros invested per year against millions saved.

‍

3 - Form teams

Quarterly sessions on new operating procedures. Concrete examples of false documents. Gamification with challenges and rewards.

Document fraud costs €105/year for every honest household. Investing now = protecting legitimate customers and improving profitability.

‍

sourcing

¹ Tessi Observatory, “France Documentary Fraud”, April 2025

² Tessi/ALFA Observatory, “Insurance sector losses”, 2025

³ Tessi Observatory, “National Fraud Survey”, 2025

4/Tessi Observatory, “Young people and documentary fraud”, 2025

452 Tessi Observatory, "+244% digital falsifications”, 2024-2025

6 Tessi/ALFA Observatory, “Falsified documents in insurance fraud”, 2025

YouGov, “France Insurance Fraud”, 2024

8 ALFA, “Key fraud figures 2022"

9 Abeille Assurances/ALFA, “Impact premiums”, 2025

10 Penal Code, articles 313-1 and 441-1

¹ Insurance Code, article L113-8

¹ ² Tessi Observatory, “Fraud Success Rate”, 2025

¹ ³ Bank of France, “Fraud evolution”, 2025

14 Finovox/ITESOFT, “Automatic detection efficiency”, 2025

‍

‍