(18)Institutions that use Gaussian or Normal-like distributions for recognising correlation within all or parts of their AMA should no longer use them in the context of their AMA as these assumptions would imply tail independence among operational risk categories, thus excluding the possibility of simultaneous occurrence of large losses of different types, an assumption which is neither prudent nor realistic. Enough time should therefore be granted for the smooth transition of these institutions to a new regime where more conservative assumptions, implying positive tail dependence, are introduced within the operational risk measurement system. Given that the implementation of these assumptions might require the modification of some key elements and the related procedures, of the AMA framework, it would be appropriate to provide two years for that transition.