The Model Law on Artificial Intelligence is a continuation of the Regulatory Institute’s popular series of model laws. The scope of the Model Law on AI applies to the development, operation and use of software that constitutes artificial intelligence or of items that use artificial intelligence. Artificial intelligence is a relatively new topic of regulation and presents a good opportunity for lawmakers to regulate in a comprehensive way, free from any legacy legislation.
In view of the different implementation capacities of jurisdictions, including those of developed economies, the following colour coding is used in this Model Law on AI:
-
Green stands for provisions which are “technically easy to implement” or “unavoidable / essential whilst being of medium difficulty to implement”; we thus recommend the green provisions for all jurisdictions.
-
The provisions marked in yellow are all of medium difficulty.
The Model Law on AI is construed in such a way that it could “stand” with the green provisions alone, but that the sections or provisions in yellow can be added as extension modules.
In the above, we refer to the aspect of technically simple and medium implementation. This parameter alone should not be the basis on which provisions are included or excluded in a particular regulation. What is more important is the overall quantitative implementation and in particular, the enforcement capacity of the jurisdiction, even very advanced technical jurisdictions should consider this carefully. Hence the following questions need to be answered:
-
Do we have the necessary (quantitative) enforcement capacities to enforce all the provisions we deem ideal?
-
If not, which provisions shall we mainly focus on in terms of enforcement, where shall we steer our resources?
As a result of this sequence of questions, even some provisions marked in green will need to be eliminated at the end of the day, and this even in technically very advanced jurisdictions. Again, we encourage the regulator to be selective.
Where two or more bits of text are in [ … ], separated by an “OR”, a choice is suggested. Short bits of text for which a choice is to be made are simply marked by a separating “ / “. Longer bits of text for which there is a choice are marked with “(Var. 1)”, “(Var. 2)”, …, “Var.” standing for “variant”. Text in “( … )” is meant to indicate how a loophole could well be filled in or gives another recommendation for use.
The purpose of the Regulatory Institute’s model laws is to facilitate the tasks of regulatory practitioners, be they working for administrations or parliaments, to improve the quality of laws by triggering more conscious choices. These model laws should be used as a toolbox, a checklist or the basis for the development of an adapted law, and optimised as such. The model laws are not intended to be used exactly as they are drafted. They try to point to important decisions to be taken by regulatory practitioners without preempting respective choices. Therefore, they often present choices, be they alternatives or add-on modules, that can be kept or deleted.
In view of that specific task, the model laws of the Regulatory Institute offer so many possibilities for differentiation that no jurisdiction will use all of them. Consequently, once a decision has been taken regarding the possibilities for differentiation to be used, the law can and should be simplified.
The Regulatory Institute’s previous model laws can be accessed below: