Artificial intelligence (AI) models used for pension allocation and annuity pricing should be included as high-risk applications under the European Union’s AI Act, the Actuarial Association of Europe (AAE) has stated.

In its response to the European Commission's Targeted stakeholder consultation on the classification of AI systems as high-risk, the AAE argued for Annex III of the act to be amended to “better reflect” the varied and evolving AI use cases in insurance and financial services.

Annex III of the act lists eight areas in which AI systems could pose a significant risk to health, safety or fundamental rights and, within each area, lists specific use-cases that are to be classified as high-risk.

One area that Annex III currently references as high risk is ‘insurance, credit scoring, and risk assessment’, but the AAE believes it could “benefit from greater granularity”.

It said this could be done by distinguishing between fully automated systems and decision-support tools that aid actuaries or underwriters. It advised the inclusion of high-impact financial decision-making systems used in insurability assessments, premium fairness evaluations, and pension/annuity pricing.

In particular, the AAE emphasised that AI tools influencing pension outcomes, particularly in public-private hybrid arrangements, should be subject to heightened regulatory scrutiny.

“These systems influence retirement income adequacy over long periods, and errors or biases in model assumptions could have far-reaching social consequences,” the AAE warned in its response.

At the same time, the AAE was careful to draw a line between modern AI systems and traditional actuarial practices. It stressed that actuarial tools such as generalised linear models, commonly used in pricing or reserving, are not typically autonomous and are developed under professional standards and oversight.

As such, they should not automatically be classified as high-risk unless deployed in a way that directly affects individuals’ rights or access to services, the AAE said.

The AI Act entered into force on 1 August 2024, aimed at creating a single market with “harmonised rules for trustworthy and human-centric AI in the EU”. However, organisations have until 2 August 2026 to comply with the regulation.

The AAE noted that although the organisation does not act as a deployer or provider of AI systems under Annex I of the AI Act, individual actuaries or their employers may do so now or in the near future.

“In particular, actuaries working in the insurance, pensions or financial sectors may contribute to the development or use of AI tools that support risk modelling, pricing or claims processes. Where such tools are commercialised, some actuaries may effectively act as providers,” it stated.

The AAE noted that actuaries are “well placed” to contribute to the responsible implementation and oversight of AI systems due to their expertise in data, modelling and risk management.

---