FRANCE: An Introduction to Insurance

Artificial Intelligence: a Tool to Step Up the Fight Against Insurance Fraud

The year 2023 has been dominated by artificial intelligence (AI). Symbolised by the launch of the ChatGPT conversational agent, AI-related innovations have been on the rise in every sector, and the insurance industry is no exception.

AI opens up new opportunities to improve various aspects of the insurance business, such as customer relations, claims management and compensation, and fraud detection.

Insurance fraud has a long history, but today it is a social phenomenon where the financial scale has a significant impact on the underwriting results of the French insurance sector. The cost of fraud in France is estimated at more than EUR2.5 billion per year, equivalent to 5% of the non-life insurance premiums.

In other words, fraud affects not only insurers but also insureds, with honest customers facing higher insurance premiums as an indirect result of fraud. In addition, the need to investigate fraud reduces the resources available to insurers to process genuine claims quickly.

The French Agency for the fight against insurance fraud (Agence de lutte contre la fraude à l'assurance – ALFA) defines insurance fraud as “an intentional act, carried out by a legal or natural person, in order to unduly obtain a profit from the insurance contract”.

Types of Fraud 

Fraud can affect all types of insurance (non-life, life, and protection or health) and can occur at any time:

(i) when the policy is taken out by providing false or incomplete information in insurance applications or answers to an insurance proposal form (intentional misrepresentation of risk);

(ii) during the policy period by making a claim based on misleading or untrue circumstances, including exaggerating a genuine claim; and

(iii) at the time of the claim (intentional misrepresentation of claim, intentional damage, etc).

Intentional misrepresentation of risk, whether made ab initio or later, is punishable under Article L 113-8 of the French Insurance Code, which provides for a severe civil penalty – ie, nullity of the insurance contract and retention of all premiums by the insurer by way damages. This provision is therefore one of the measures available under French law to punish insurance fraud.

In addition, voluntary damage, qualified as intentional fault or wilful misconduct on the part of the insured, is subject to a public policy legal exclusion (Article L 113-1 of the French Insurance Code).

On the other hand, there is no provision specifically sanctioning deliberate misrepresentation of a claim. It is therefore up to the policy to provide a forfeiture of cover in such cases. The implementation of this clause is subject to compliance with a certain contractual formalism (Article L 112-4 of the French Insurance Code states that “policy clauses stipulating [...] forfeiture of cover are only valid if they appear in very visible characters” and proof of bad faith on the part of the insured.

Under French law, insurance fraud is also punishable under criminal law as attempted deceit or deceit.

Insurance fraud is difficult to detect because it does not have the characteristics of an obvious offence. Moreover, insurance fraud is constantly evolving, driven by the technology available to fraudsters. In recent years, electronic fraud has become more common, as increasingly more insurance transactions are conducted online.

Combating Fraud 

The insurance industry is proactive in combating fraud in a number of ways, including:

i) dedicated investigation groups;

ii) co-operation with law enforcement agencies;

iii) provision of specialised anti-fraud training;

iv) use of technology and data analysis (including anti-fraud databases); and

v) information campaigns, etc.

In France, ALFA was set up by insurers in 1989 to investigate fraud and to collect evidence. ALFA also aims at promoting anti-fraud activities by creating tools to help the industry fight fraud.

From 2015, claims handlers began to be supplemented by digital tools. Insurers had moved from simple manual queries and reports to predictive algorithms that can cross-reference up to 100 data points per claim to detect fraud.

In 2019, ALFA launched a detection tool to help the market fight organised crime in motor insurance. French insurers submit their data on contracts and claims to a third party, which supplements it with data from other sources (such as expert reports and data from third parties) and then generates alerts drawing insurers’ attention to potential cases of fraud.

However, these tools for detecting cases exposed to the risk of fraud remain insufficient, with the rate of frauds detected at only 15%. New technologies, and AI systems in particular, can now help meet the insurance sector’s need for industrialisation and enhanced performance.

In April 2021, the European Commission proposed the European Union’s first regulatory framework for AI.

On 14 June 2023, the European Parliament voted in favour of a new version of the draft regulation, on which provisional political agreement was reached between the European Parliament and the Council on 8 December 2023 (the “Artificial Intelligence Act”), and which now needs to be completed through several technical works.

The Artificial Intelligence Act is based on a risk-based approach to the use of AI systems, with four levels of risk practices:

i) prohibited;

ii) high risk;

iii) limited risk; and

iv) minimal risk.

According to this text, AI systems used to detect fraud in the provision of financial services should not be considered high risk.

The fight against fraud appears to be one of the areas most likely to benefit from these advances. Indeed, AI fraud detection can be used to identify anomalies in individual claims and search for fraud patterns across all claims, which could help reduce payments to fraudulent and illegal claims as well as overall costs.

Performance metrics (accuracy, recall, precision, etc) depend on the nature of the data used and the intended application of the AI. For example, in classification AI systems used in fraud detection, insurance firms should decide whether the objective is to maximise the prediction accuracy (number of fraudulent claims detected), or to reduce the number of “false positives” (legitimate claims wrongly labelled as fraudulent) or “false negatives” (claims labelled as legitimate which ultimately are fraudulent).

AI makes it possible to combat insurance fraud from the underwriting stage by:

i) detecting potentially fraudulent identities and applications (underwriting risk detection);

ii) automatic recognition of documents; and

iii) more effective reconciliation of identity data (if the same person presents several pieces of identification for several transactions), etc.

AI can also be used to identify potential cases of fraud in the context of claims declarations with a very high degree of accuracy and detailed contextual indications, facilitating the work of anti-fraud teams by revealing:

i) recurring patterns of action by the policyholder;

ii) inconsistencies in the declaration of claims;

iii) unusual recurrences of events; or

iv) the provision of identical RIBs in different files (claims fraud detection).

However, the European Insurance and Occupational Pensions Authority (EIOPA) considers that fraud analytics techniques powered by AI systems should not be able to classify any customer or transaction as “fraudulent” without the involvement of relevant staff, both at the pre-contractual stage and during a claim.

Outlook 

In the years to come, we should see more insurtech start-ups appear. These companies, which take advantage of technological advances to transform and improve the insurance sector, mainly use cutting-edge technologies such as AI, machine learning, big data analysis and blockchain to streamline processes, improve the customer experience, and introduce new insurance products and services.

Some of these start-ups specialise in the detection and prevention of fraud in the insurance sector, and use AI to detect potentially fraudulent entities and applications prior to account-opening, and to identify potential claims fraud with extreme accuracy and detailed contextual guidance to empower investigators. These start-ups are already proving so successful that some have already become unicorns.