We already have rules for AI systems
The European Commission has published a draft for an
AI (Artificial Intelligence) Regulation, since it does not consider current laws and regulations are sufficient for regulating AI systems. The approach of the draft is to choose human-centered AI. Developers of AI systems must assess for themselves in which of the four risk groups their system falls. The higher the risk of an AI system, the higher the requirements for that AI system. It will still take years before the AI Regulation will come into effect. There is also good chance that the draft will be adjusted in the mean time.
What does this draft mean for the AI systems that are currently being developed or used? Are there no rules which apply to them?
For several types of AI systems, there are already laws and/or regulations that must be complied with. For example:
- Medical Devices Regulation: for AI systems in medical devices;
- Constitutions + Human Rights Treaties: for protection of fundamental rights such as freedom of speech, privacy and self-determination;
- General Data Protection Regulation: when personal data is processed;
- Product safety regulations: when an AI system causes injury;
- Consumer protection: when information obligations arise from these regulations;
- Codes of conduct: when rules (code of conduct) have been established in a sector for AI systems, and
- Contracts: when parties have established rules for AI systems in an agreement.
In short, even without a specific AI regulation, there are already
laws and regulations that an AI system must comply with. For AI developers, this means that they must check which laws and regulations their AI system must comply with and whether they are complient with these rules. If they are not, a customer/client may later claim that the AI system did not meet the legal requirements. This can lead to claims for damages. As a user/client for the development of an AI system you will have to ask similar questions. This applies to supervisors of parties using an AI system.
General information about legal aspects of AI can be found on our
knowledge platform LegalAIR.
AI compliance assessment
For this reason, it is wise for both developers and users of AI systems to determine with an AI compliance assessment which laws and regulations apply to their AI system and whether they are compliant. If they are not met, timely measures can be taken to become compliant.
For more questions about this, or for guidance on the performance of an
AI compliance assessment, please contact Jos van der Wijst.