The Artificial Intelligence Act – update
Over the past few years, the spread of artificial intelligence techniques into various areas of life has sparked off debate on the need for regulation in this area. As part of the European Union’s digital strategy, the EU Commission presented a proposal for a basic regulation to address this issue on 21 April 2021 – the Artificial Intelligence Act.* My articles from 3 February 2022 and 22 February 2022 provided an overview of the system and regulations.
Compromise text from the European Council
The EU Commission’s proposed regulation will continue to be negotiated in a trilogue with the institutions involved in the legislative process (the European Parliament and the European Council). The bodies examine the proposal and suggest amendments or additions. Based on the EU Commission’s proposal, the European Council has done just that and published a compromise text on 29 November 2021.**
And it’s a tough one, especially from the insurance industry’s point of view. But let’s take one thing at a time.
Requirements for high-risk systems
Just as a quick reminder, the European Commission’s proposal uses a risk-based approach to classify AI systems into four categories. AI systems are allocated to one of the four categories, determining how extensive the regulation is applied. The focus is on high-risk systems, which are subject to comprehensive rules. These include specific risk management, data governance, detailed technical documentation and record-keeping requirements, transparency regulations, human oversight and specific requirements for accuracy, robustness and cybersecurity.
In addition, there are far-reaching obligations for providers of high-risk systems. These include the establishment of a quality management system, the obligation to carry out regular conformity assessments as well as extensive information obligations towards the relevant authorities.
Classification of AI systems in insurance as high-risk systems
Criteria for classifying systems as high-risk are listed in Annex III pursuant to Article 6, where a total of eight specific areas are identified. From the insurance industry’s perspective, the only relevant specific area, group five, could be: ‘Access to and enjoyment of essential private services and public services and benefits’. However, the European Commission’s proposal only refers to AI systems in public authorities, banks, emergency services and fire brigades. We can’t see anything about insurance here!
This is changing with the European Council’s compromise text, which proposes adding a fourth category to group five. The gist of the text is: ‘AI systems to be used for setting insurance premiums, risk assessments and damage assessments.’ This would classify numerous AI applications in insurance companies as high-risk systems, making them subject to extensive regulation!
Definition of AI systems in the Artificial Intelligence Act
This may not only affect the use of AI systems in the conventional sense. Article 3, paragraph 1 of the European Commission’s proposal provides a very broad definition of AI systems and specifies the techniques and approaches in Annex I. This also covers logic and knowledge-based systems, as well as statistical approaches. Accordingly, in the future, current procedures for risk assessment, claims evaluation and premium determination, which we generally do not consider to be AI-based, could also fall under the above regulation.
However, the European Council’s compromise text clarifies the definition and narrows down what an AI system is considered to be. Traditional software systems, which are not normally regarded as AI, should therefore be excluded from the scope of applicability.
The GDV’s position
On 26 January 2022, the GDV published an opinion***, in which it explains why it considers the above expansion of group five to be unjustified. The GDV refers to the existing extensive and strict form of regulation in the insurance industry as well as the basic principles of risk-based premium setting and risk assessment – regardless of whether AI systems are used for this purpose or not. Finally, the GDV adds that decisions in the insurance industry are always verifiable and reversible. Customers always have the right to demand a further review (including a supervisory review) of decisions in the insurance business, regardless of whether AI systems are used or not.
How it stands
The European Council’s compromise text on the Artificial Intelligence Act may create a completely new situation for the insurance industry. While the original proposal for AI systems had almost no impact on insurance companies, the European Council’s compromise proposal provides for extensive regulation.
However, nothing has been set in stone yet! The pendulum may still swing one way or the other until adoption, which is currently expected in autumn 2022. We will keep you informed.
*) Proposal for a REGULATION OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL LAYING DOWN HARMONISED RULES ON ARTIFICIAL INTELLIGENCE (ARTIFICIAL INTELLIGENCE ACT) AND AMENDING CERTAIN UNION LEGISLATIVE ACTS
**) Proposal for a Regulation of the European Parliament and of the Council laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) and amending certain Union legislative acts – Presidency compromise text
***) Position Paper of the German Insurance Association (ID Number 6437280268-55) on the EU regulatory framework for artificial intelligence, 26 January 2022