Much AI (regulation) about nothing?

Earlier this week, the European Commission released the draft of the proposal of the regulation laying down harmonised rules on artificial intelligence (“AI”). This proposal has livened up an already scintillating discussion over artificial intelligence. The proposal is one of the first comprehensive frameworks and it painstakingly lists the general obligations of AI providers. It includes systematic reviews of the activities, documental compliance and the creation of a European Board for AI. The proposal does not define what AI is, but it simply indicates a series of techniques: machine learning, logic-, statistical- and knowledge-based approaches.

Moreover, it details cases in which the AI shall be prohibited (e.g., when it can cause physical and/or psychological damage) and how to mitigate the risks in case of high risk activities undertaken by an AI. The Commission chose to give the Authorities anti-competition-like sanctions: the maximum fine for non-compliance could be between 2% and 6% of the offender’s annual global turnover, according to the violations. The proposal has received strong criticism from privacy lawyers, for being either a burden (it is not easy to prove a link between an AI and psychological damage) or ineffective  (large corporations may not be sufficiently deterred by the fines at stake), but at the same time has been criticised by companies for being too stringent. The proposal is currently open for feedback, thus, if you think you can contribute, do not hesitate to raise your voice.