News

The EU AI Act: From Understanding the Rules to Taking Informed Action

‹ News

The EU AI Act: From Understanding the Rules to Taking Informed Action

The legal requirements governing artificial intelligence are no longer a distant concern. Some provisions of the EU AI Act already apply, while the main requirements for high-risk AI systems will take effect later. Businesses therefore have good reason to consider not only the opportunities offered by AI, but also the obligations that come with its use.

In an ideal world, advances in technology would always translate into improvements in people's lives. Yet technological progress and human well-being do not necessarily go hand in hand. How, then, can we harness the benefits of technology without allowing people's interests to be overlooked in the pursuit of innovation? This is precisely the question that the European Union's AI Act seeks to address through a legal framework.

“But We Don't Develop AI”


That alone does not place a company outside the scope of the AI Act. The Regulation applies both to AI developers and to businesses that use AI in their operations.

However, not everyone is subject to the same obligations. Their scope depends on the company's role and the risks associated with the system. This risk-based approach is central to understanding the AI Act: the greater the risk, the more extensive the requirements.

The starting point for assessing risk is not so much the AI system itself, but rather what its outputs may affect.

How Can Businesses Determine Which Requirements Apply to Them and What They Need to Do?


Consider recruitment, for example. AI can help improve the wording of a job advertisement. Asking AI to decide which candidates should be invited for an interview is an entirely different matter.

In the latter case, a person's opportunity to obtain employment is at stake. This is why AI systems intended to evaluate job applicants are classified as high-risk under the Regulation. However, this does not mean that every use of AI in human resources automatically triggers the same set of requirements.

As is often the case in law, there are exceptions. For example, a system performing a narrowly defined technical or preparatory task may fall outside the requirements applicable to high-risk AI systems, provided that the conditions for the exception are met.

It is therefore worth considering, already when selecting or developing an AI solution, what task the system is actually intended to perform. In some cases, the same objective can be achieved with a simpler solution and a lighter regulatory burden.

What Should Businesses Be Doing Now?


A practical starting point is to take the following steps:

1. Map your AI systems. Identify which AI systems your company uses and/or develops.
2. Determine your company's role. Establish whether your company acts as a provider or a deployer of each system.
3. Identify prohibited AI practices. Check whether any of your AI systems involve practices classified as posing an unacceptable risk.
4. Assess high-risk systems. Determine whether any systems qualify as high-risk and what requirements apply to them.
5. Review transparency obligations. Establish whether and how you are required to inform individuals about the use of AI.
6. Ensure AI literacy among employees. Check whether employees understand the risks associated with AI and know how to use AI tools responsibly and within permitted boundaries.
7. Verify third-party solutions. When using an AI solution developed by another company, check whether it complies with the applicable requirements. Remember that using a third-party solution does not relieve your company of its own obligations.

Importantly, the AI Act does not replace other existing legal requirements. Businesses must still comply with applicable rules on personal data protection, intellectual property, consumer protection and other relevant areas of law.

These issues should be carefully considered before integrating an AI solution into a company's business processes.

This article was written by Karl Erik Esko, Attorney-at-Law and Co-Head of RASK's Technology Law practice.