As almost everyone will be aware, the EU has launched a draft plan to regulate 'AI'. This is, of course, only the beginning and the one certainty is the end result is highly unlikely to resemble the final product.

Although there are significant policy questions regarding the evolving development and implementation of 'AI' systems, and it is laudable that the EU is seeking to grapple with those, the question arises is a general law likely to be so wide as to be ineffective. The first rule of law-making is, or at least ought to be, whether the law will actually have the desired effect. However, that requires laws to be targeted, and to address specific issues.

The question here is at what level of abstraction can laws on 'AI' be effective? The best example I've seen to explain that is from @benedictevans, who comments that we do not have a 'car law', but rather many laws that cover things like speeding, parking fines, emissions, taxation, driving under the influence and so on. One therefore has to wonder if, to be effective, any regulation of 'AI' should take place at that lower level of abstraction, or at least on a sector-by-sector basis. For example, the implications of AI in the healthcare context are likely to vary greatly from those in the finance sector. However, only time will tell.