Regulation of AI in Insurance: Evolution or Revolution?
Debate on regulation of AI in insurance focusing on balancing benefits and risks. Discusses the EU AI Act, possible regulatory approaches, and implications of AI in the insurance industry, addressing socio-economic benefits, ethical concerns, and the impact on mutualisation principles.
Download Presentation
Please find below an Image/Link to download the presentation.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author. Download presentation by click this link. If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.
E N D
Presentation Transcript
Regulation of AI in insurance: evolution or revolution? Prof. Em. Karel Van Hulle KU Leuven and Goethe University Frankfurt AZN Conference Ljubljana, 11 September 2023
Why do we have this debate? AI will change our lives, for instance by improving healthcare, increasing efficiency, contributing to climate change mitigation and adaptation, .(benefits) AI entails a number of potential risks, such as opaque decision- making, gender-based or other kinds of discrimination, intrusion in our private lives or being used for criminal purposes, (risks) Regulation of AI should promote the uptake of AI so as to maximise the benefits and address the risks associated with certain uses of this new technology in order to exclude or reduce these risks Existing or planned regulatory initiatives therefore refer to a responsible use of AI or to trust and excellence in using AI, following a risk based approach Will we strike the right balance? Prof. Karel Van Hulle - KU Leuven and Goethe University Frankfurt 2
Regulatory approach The EU AI Act, which is a horizontal regulation, adopts a risk based approach, distinguishing between AI systems that pose an unacceptable risk (prohibition); a high risk (subject to a set of new rules with an ex-ante conformity assessment and a CE marking); a limited risk (transparency); a low or minimal risk (no obligations) Questions: Can the list of prohibited AI systems be enlarged by the EC? Can the treatment of high risk AI systems be left to self-assessment or should there be a conformity assessment by an independent entity prior to their deployment? Should there be an explicit right of redress for individuals? o o o o o o o Prof. Karel Van Hulle - KU Leuven and Goethe University Frankfurt 3
AI in insurance AI may provide socio-economic benefits for insurance by expanding the scope of risk pooling, reducing the cost of risk pooling or mitigating and preventing risks AI may also make risk-sharing obsolete, tilting the information asymmetry between customer and insurer too far in favour of the insurer, resulting in discrimination and exacerbating social exclusion, thereby calling into question the very principle of mutualisation on which insurance and its social pact are founded Insurance is based upon trust and a wrong use of AI might damage this trust Insurance is also about ethics, combining the separate interests of individuals seeking insurance cover or being insured, the pool of insured risks and the insurer who manages the pool Prof. Karel Van Hulle - KU Leuven and Goethe University Frankfurt 4
Guiding principles In order to promote responsible AI in insurance and to ensure ethical and trustworthy AI in insurance, the following governance principles have been proposed by experts from the insurance industry (Geneva Association) and by insurance supervisors (EIOPA): Fairness and non-discrimination (diversity) Transparency and explainability Human oversight Data governance and record keeping (privacy/accountability) Robustness and performance EIOPA adds the principle of proportionality which should be applied by insurers and insurance distributors in the impact assessment of the governance principles o o o o o Prof. Karel Van Hulle - KU Leuven and Goethe University Frankfurt 5
Governing principles in practice (1) Fairness and non-discrimination: similar risks must be treated similarly, the premium should not be based on irrelevant factors that an individual cannot influence, reasonable efforts must be made to monitor and mitigate biases from data and AI systems; Transparency and explainability: data used in AI models should be communicated transparently to consumers and consumers must be made aware that they are interacting with an AI system and its limitations Human oversight: the organisational structure of insurers should assign and document clear roles and responsibilities for the staff involved in AI processes Prof. Karel Van Hulle - KU Leuven and Goethe University Frankfurt 6
Governing principles in practice (2) Data governance and record keeping: data used in AI systems must be accurate, complete and appropriate and the same data governance standards should be applied to data obtained from internal or external sources; data should be stored in a safe and secured environment Robustness and performance: AI systems should be robust, both when developed in-house or outsourced to third parties, taking into account their intended use and the potential to cause harm; the calibration, validation and reproductibility of AI systems must be done in a sound manner that ensures that the AI system outcomes are stable overtime and/or of a steady nature Prof. Karel Van Hulle - KU Leuven and Goethe University Frankfurt 7
Evolution or revolution? The application of the governing principles brings us back in known territory: risk management and governance Fairness and non-discrimination: fair treatment of customers, non- discrimination, product oversight and governance (IDD) Transparency and explainability: provision of objective information about the insurance product in a comprehensible form (IDD) Human oversight: insurers must establish an adequate transparent organisational structure and create key governance functions (SII) Data governance and record keeping: data quality requirements for the calculation of technical provisions and for internal models (SII) Robustness and performance: robustness and performance in the calculation of technical provisions (SII) and new operational resilience and IT security requirements (DORA) Prof. Karel Van Hulle - KU Leuven and Goethe University Frankfurt 8
Concluding remarks There still is no commonly agreed definition of AI AI has been in existence for some time and is likely to develop further into not yet known territories There is no need to embark on a new series of rules and regulations that specifically address AI The prohibition of harmful systems of AI is necessary in order to protect people s safety, livelihoods and rights The AI governance principles developed by EIOPA for ethical and trustworthy AI in the European insurance sector are a useful tool to integrate AI concerns in the already existing legal framework for insurers and insurance distributors In a fast developing area, it is difficult to find the right balance between protection and freedom of operation Prof. Karel Van Hulle - KU Leuven and Goethe University Frankfurt 9
