Professor Susanna Lindroos-Hovinheimo from the University of Helsinki has listed three points that everyone should be aware of regarding the proposed AI legislation:
1. AI products are the subject of the regulation
The Commission primarily considers the regulation to consist of product regulation. The legislation is supposed to limit the systems that can be used in and imported into the EU single market. The goal is to increase the confidence of citizens in these products, i.e. AI systems within the EU, and to create as functional and free single market as possible.
Contrary to the much discussed GDPR (General Data Protection Regulation), the regulation of AI is not based in human rights or their safeguarding. However, these issues are considered as part of the pack.
2. Regulation based on risks
The legislation would classify AI systems into systems that are forbidden, high-risk, or low-risk. The definitions would be made according to what kind of threat the systems could pose to individuals and society.
According to the proposal, banned systems would include e.g. those that can modify human behavior in a negative way, abuse the weakness of any group of people, or could rank people.
High-risk systems are e.g. traffic infrastructures, components related to security (such as robot surgery), and public administration systems that hold great value for people.
Most of the regulation would pertain to high-risk systems. For them, the Commission suggests certain requirements that the systems have to fulfill, or else they could not be developed for the EU market.
3. One set of AI rules for the whole European Union
The new rules would be binding as such in the whole European Union. This is why the Commission is suggesting this legislation to function as a regulation.
This would not leave EU member countries much leeway for their own legislation, i.e. in practice everything connected with AI would be regulated at the EU level.
The goal is to have the same AI rules applied in all member states, which would also make it easier to control the products. The Commission would like to see a future with one shared market for AI products across the whole European Union.
What happens next?
The proposal from the Commission is just a suggestion so far. It will be processed in the same way as all EU legislation does. After that, the member states will negotiate the contents of the regulations. The EU Parliament also has to approve the regulation before it can come into use.