Good morning ladies and gentlemen,
I am very happy to welcome you today and to thank you for joining us to learn more about the use of Artificial Intelligence in the insurance sector, and more particularly about AI governance.
We organized this public event to engage with you on this important topic which has a direct impact on the present and the future of the insurance industry.
At EIOPA we want to continue to closely cooperate with stakeholders ensuring that the use of AI in the European insurance sector, is done in an ethical and trustworthy manner that brings benefits for the industry, consumers, and society as a whole.
Digital transformation and value chain
Artificial Intelligence plays an essential role in the digital transformation of the insurance industry. This can be illustrated by the multiple use cases that can be found through the entire insurance value chain.
In today’s event we will have the presentation of three different, interesting and very relevant use cases: the use of AI in motor insurance telematics, the use of AI in liability allocation in the event of an accident, and the use of AI in natural catastrophes risk modelling.
But the use of AI in insurance does not stop there. There are many other AI use cases in every area of the insurance value chain, including both consumer-facing and back-office operations.
Opportunities and risks
AI brings many benefits for the insurance industry and consumers; indeed, the benefits arising from AI in terms of prediction accuracy, automation, efficiency and speed are remarkable.
For example, AI algorithms have managed to “convert” texts and images into data enabling insurance undertakings to quickly assess claims, dramatically speeding up claims processing times, enabling more accurate risks assessments and more efficiently fight against fraud.
But AI also brings relevant challenges and risks for consumers, two of which will be directly addressed in today’s event: the issue of explainability of AI systems, or ways to overcome the so-called “black-box” effect of some AI systems, and the one of AI fairness and how to address algorithmic biases.
The use of mathematical models is not new for the insurance industry and the sector has already extensive experience on data analytics processes and governance and risks management measures around them.
But AI has some unique characteristics that need to be incorporated into existing governance frameworks, especially in AI use cases that have a potential high impact on consumers and undertakings themselves.
This is precisely the focus of today’s event; to learn from experts with hands-on experience about the use of AI and the governance and risk-management measures that they are implementing in their respective organizations.
At EIOPA we believe that ethical and trustworthy AI systems are achieved not by a single governance measure (e.g. explainability, human oversight, record keeping etc.), but rather by a combination of all them, following a risk-based approach and adapting them to the concrete AI use case at hand.
This was indeed one of the key findings of EIOPA’s thematic review on the use of Big Data Analytics in motor and health insurance that we conducted in 2018-2019.
The thematic review also showed that at that time already one third of the insurance undertakings in the EU were actively using AI, and another third of them where at a proof-of-concept / experimentation stage.
Next year we plan to issue a digitalization market monitoring survey that among other things will allow us to update these figures, although we expect that the uptake of AI in the industry will have continued to increase in view of the benefits.
But again, there are also challenges and for this reason EIOPA convened an Expert Group on digital ethics in insurance, which in 2021 developed an AI governance principles report providing guidance to the sector.
In the years to come EIOPA will continue this cooperation with stakeholders and providing further guidance, at European. Also, at international level through IAIS and its Fintech Forum, which is chaired by EIOPA, we will continue to share practices and identify what regulatory steps are currently taken
We will also seek to gather a better understanding of the implications of the trend towards increasing data-driven business models from a financial inclusion perspective.
But in doing so EIOPA is not isolated, and we need to consider the developments that are taking place at cross-sectorial level in the EU such as the AI Act.
Let me make a few remarks about the AI Act before I conclude.
AI systems do not operate in an unregulated world. A number of legally binding rules at international, European and national level already apply or are relevant to the development, deployment and use of AI systems today. Legal sources include, but are not limited to: EU primary law (the Treaties of the European Union and its Charter of Fundamental Rights), EU secondary law such as the Insurance Distribution Directive, Solvency II Framework, the General Data Protection Regulation or GDPR, the Product Liability Directive, as well as the Commission’s recently proposed AI Liability Directive, anti-discrimination Directives, or the Unfair Commercial Practices Directive, the UN Human Rights treaties and the Council of Europe conventions (such as the European Convention on Human Rights), and numerous EU Member State laws.
In particular for Solvency II Directive for example, Article 41 (1) of Solvency II Directive requires “insurance and reinsurance undertakings to have in place an effective system of governance which provides for sound and prudent management of the business.”
With regard to the Insurance Distribution Directive, the requirement that insurance distributors act in the best interests of their customers is in line with the principle of fair processing of data contained in the GDPR. Likewise, product oversight and governance requirements encompass fair and ethical use of data.
Existing legislation should indeed form the basis of any AI governance framework, but the different pieces of legislation need to be applied in a systematic manner and require unpacking to assist organisations understand what they mean in the context of AI.
Furthermore, an ethical use of data and digital technologies implies a more extensive approach than merely complying with legal provisions and needs to take into consideration the provision of public good to society as part of the corporate social responsibility of firms.
In addition, the European Commission’s legislative proposal for an Artificial Intelligence Act will provide a cross-sectorial legal framework for the use of AI in the European Union.
Since technology fosters connections and interconnections, we shouldn’t be surprised that regulatory frameworks are moving from being sector specific to horizontal.
This can be a good thing, and so when it comes to the AI Act, and at EIOPA we support the European Commission’s risk-based approach of the AI Act.
But we also feel that it is very important for the specificities of the insurance sector to be fully taken into consideration and to recognize that sectoral legislation is already addressing the risks of the use of AI and mandating supervisors to act when needed.
Therefore EIOPA does not support the inclusion of insurance AI use cases in the list of high-risk applications of the AI Act at this moment in time; further specification of the AI framework should be dealt with by sectorial legislation, building on the already existing sectorial governance, risk management, conduct of business and product oversight and governance requirements.
Insurance supervisors are already supervising AI risk and they know very well how their sector operates. We therefore believe that national supervisors – together with EIOPA – should remain responsible for the further development and implementation of further regulation and supervision of the use of AI in the insurance and pensions sector. Of course this will be done in close cooperation with the proposed AI Board to ensure a consistent but proportionate approach.
Finally, the definition of AI should not cover mathematical models that have traditionally been used and regulated in the insurance sector, including internal models. In our view, the definition of AI should be narrower and focus on AI systems that have distinctive features, such as machine learning approaches.
As technical advisory body to the EU Institutions, we will continue our regular dialogue with the European Parliament, the Council and the Commission and to support the policy work at EU level, while doing our best to make sure that the balance is right for the insurance industry.
To conclude I would like to say that it is good to see that this event has attracted a lot of attention, which shows the relevance for the sector of the topic that we are about to discuss.
I wish you all a productive and efficient meeting and that you will leave the meeting with some useful insights that we can apply in the our future work.
Let me also warmly thank our distinguished and experienced speakers who have made some time in their busy agenda to share with use their know-how and experience.
Thank you very much.
- Publication date
- 15 December 2022