Insurance and Pension Denmark (I&P Denmark) has warned against “Danish over-implementation” of the European Union’s artificial intelligence regulation.

Although it welcomes the new EU rules, it cautioned that over-implementation could lead to stricter rules than necessary, and result in unnecessary administration.

The EU's comprehensive legislation on the use of artificial intelligence (the AI Regulation) has just been published and will officially come into force on 1 August.

Commenting, I&P Denmark director of insurance and digitalisation, Thomas Brenøe, said: “The regulation is not entirely concrete in its text, as there is a lot of room for interpretation. This means that companies and authorities will have to wait for the guidelines and standards that will be necessary in practice to know what changes and processes to introduce to comply with the regulation.

"Unfortunately, we often find that Denmark becomes the EU’s puppet, meaning that the rules are given an extra twist – perhaps to be on the safe side. Unfortunately, this is a very unfortunate tendency, as it means that we end up with stricter rules than the politicians actually wanted.”

Whether the implementation succeeds in maintaining the desired balance between risk and requirements – a balance that the European Parliament and the Council have worked hard to maintain in the political process – will be crucial for the regulation, according to I&P Denmark.

"If the content and scope of the requirements outweigh the risk they are intended to address, it is in fact unnecessary administrative burdens, which instead of contributing to more responsible AI use, risks discouraging companies from using the technology altogether," Brenøe added.

The Danish pension association believes that the regulation has struck a sensible balance between risk and regulation. The riskiest applications of AI are banned – such as AI systems that use deliberately manipulative or deceptive techniques.

There are also a number of requirements for high-risk scenarios, such as using AI for emotion recognition or risk assessment in financial scenarios, while the use of low-risk AI is largely exempt from requirements.

In addition to maintaining the desired balance in the existing requirements, the upcoming guidelines must also define requirements that are not described in the regulation.

“When the AI Inspectorate visits a company in the future, they will be expected to show that they have considered whether their use of artificial intelligence falls under the regulation – regardless of whether it is high or low risk. However, the regulation does not describe how this should be done," Brenøe explained.

I&P Denmark is concerned that companies may be required to make detailed overviews of all AI systems with associated risk assessments and detailed justifications for why the individual systems fall within or outside the regulation.

“The political process in the EU has worked hard to maintain a risk-based approach. This is reflected in the fact that the regulation sets very few requirements for the use of low-risk artificial intelligence. When the regulation comes into force, it is important to maintain this approach and ensure that companies and authorities do not have to submit extensive reports to show that they are not covered by the regulation," Brenøe concluded.

---