EU AI Act: what it means for actuaries

The importance of artificial intelligence (AI) is growing rapidly in an increasingly digitalised world – particularly for actuaries. They face the challenge of navigating a dynamic environment shaped not only by new technologies but also by complex legal requirements. The European Union’s AI Act , adopted on 13 March 2024 and in force since 1 August 2024, will fundamentally change how AI is used, creating new opportunities and risks for the insurance and financial sectors. Companies must implement the new requirements by 1 August 2027. Non-compliance may result in substantial fines of up to 7% of annual turnover or €35 million.

The AI Act defines AI systems as machine-based systems with varying levels of autonomy that can make decisions and influence physical or virtual environments. Systems are categorised according to their risk, with applications that process personal data considered particularly high-risk. The associated risks – such as data protection, biased behaviour and security concerns – are to be addressed through principles such as fairness, transparency and accountability. For actuaries, this means harnessing the value of AI while minimising the associated risks. Responsible use of AI is crucial to long-term success and requires continuous professional development and adaptability.

Actuaries are called upon to develop their expertise in areas such as risk management and data analysis to use AI responsibly while promoting efficient and innovative solutions. Expertise in computer vision, robotics, machine learning and programming languages such as Python, R and SQL opens up new opportunities for them. This is accompanied by new areas of responsibility, such as ensuring data quality and model transparency, as well as scaling AI systems operationally.

Responsibilities for actuaries

Key new areas of responsibility for actuaries include:

1. System inventory and risk mitigation: Establishing an inventory to track AI systems (e.g. through logging) and their regulatory relevance is essential, particularly for applications involving risk. This involves using risk mitigation strategies that align with governance frameworks such as Solvency II and comply with GDPR requirements.

2. Interdisciplinary collaboration: Actuaries must work closely with data scientists, compliance teams and experts from various disciplines to ensure transparency, reliability and efficiency. Successful AI adoption depends on a well-trained workforce and structured risk management.

3. Explainable AI (XAI): By distinguishing between developer-oriented and user-oriented models, actuaries must ensure that AI outputs are understandable and traceable. Tools such as SHAP and LIME support technical debugging, while accessible explanations are essential for non-technical stakeholders.

In conclusion, it is worth reiterating that AI is becoming increasingly important for almost all companies (see the EIOPA study of April 2024: Report on the digitalisation of the European insurance sector) and that the associated regulations are necessary to ensure security and data protection for everyone. The AI Act not only introduces regulatory requirements but also offers actuaries the opportunity to actively shape the future of AI and promote its safe and ethical use.