The North Atlantic between Newfoundland and Ireland is practically the only region of the world that has defied global warming and even cooled. Last winter there even was the coldest on record – while globally it was the hottest on record. Our recent study (Rahmstorf et al. 2015) attributes this to a weakening of the Gulf Stream System, which is apparently unique in the last thousand years.
...
Climate models have long predicted such a slowdown – both the current 5th and the previous 4th IPCC report call a slowdown in this century “very likely”, which means at least 90% probability. When emissions continue unabated (RCP8.5 scenario), the IPCC expects 12% to 54% decline by 2100 (see also the current probabilistic projections of Schleussner et al. 2014). But the actual past evolution of the flow is difficult to reconstruct owing to the scarcity of direct measurements. Therefore, in our study we use data on sea surface temperatures in order to infer the strength of the flow: we use the temperature difference between the region most strongly influenced by the AMOC and the rest of the northern hemisphere.
...
What is new is that we have used proxy reconstructions of large-scale surface temperature (Mann et al, 2009) previously published by one of us (study co-author and RealClimate co-founder Mike Mann) that extend back to 900 AD (see “What we can learn from studying the last millennium (or so)”) to estimate the circulation (AMOC) intensity over the entire last 1100 years (Fig. 3). This shows that despite the substantial uncertainties in the proxy reconstruction, the weakness of the flow after 1975 is unique in more than a thousand years, with at least 99 per cent probability. This strongly suggests that the weak overturning is not due to natural variability but rather a result of global warming.
...
If our analysis is correct, then this indicates that climate models underestimate the weakening of the Atlantic circulation in response to global warming – probably because the flow in these models is too stable (see Hofmann and Rahmstorf 2009). Although these models predict a significant weakening for the future, they do not suggest this as early as the observations show it (see Fig. 2 of our paper). That the real flow may be more unstable than previously thought would be bad news for the future.

http://www.realclimate.org/index.php/archives/2015/03/whats-going-on-in-the-north-atlantic/