The insurance industry offers the best conditions for success in the new, digital (data) world. Dealing with data and models has been part of the core business of insurers for a long time now. In addition, many companies are aware of the potential that modern artificial intelligence (AI) and machine learning (ML) methods have to offer. AI/ML can help to calculate prices in a risk-adequate way and to assess risks more precisely. AI/ML can be considered a success factor on the costs and distribution side, too. Processes can be optimised thanks to new technologies (e.g. automated claims processing), and new cross and upselling opportunities are emerging as well.

The primary task of supervisors remains unchanged

All those opportunities that insurers (will) seize – partly because this is in their own interest – do not change the primary task that supervisors must fulfil. They must particularly ensure that the interests of both individual policyholders and the pool of policyholders are protected, including in the new, digital (data) world. To ensure that this goal can continue to be achieved, BaFin is following a risk-sensitive supervisory approach, which has proven to be effective: AI/ML methods should be supervised in a technology-neutral and risk-adequate manner.

Additional requirements only if there are additional risks

However, it would not be expedient to address the perceived need to intensively supervise the use of such methods based on vague definitions of AI/ML. Additional requirements are appropriate only if there are additional risks. General regulatory approaches which seek to make a distinction between AI/ML and conventional methods using an exhaustive and yet relatively vague list of algorithms are therefore neither appropriate nor practical for the insurance sector. This would also result in significant legal uncertainty, which could even hamper the use of innovative technologies in the insurance sector.

Instead, it would make more sense to apply the regulatory tools already in place to AI and ML – in a risk-adequate manner, as described above. The regulatory framework that is currently applicable already offers the amount of flexibility needed in order to deal with financial innovations in a risk-adequate – and thus proportionate – manner based on their characteristics. Of course, new challenges will arise, such as the high level of flexibility and sometimes high-frequency adaptability of AI/ML methods. Questions surrounding data quality and data governance are also becoming increasingly relevant. These questions need to be answered, particularly in relation to the specific area of application.

Well-prepared for the future

BaFin is already supervising the complex actuarial methods used by insurers with a technology-neutral and risk-based approach. This makes us confident that we are prepared for new developments, too.