First the Economist, now Reuters, has now noted the long period over which the consensus has been incorrect:

OVER the past 15 years air temperatures at the Earth’s surface have been flat while greenhouse-gas emissions have continued to soar. The world added roughly 100 billion tonnes of carbon to the atmosphere between 2000 and 2010. That is about a quarter of all the CO₂ put there by humanity since 1750. And yet, as James Hansen, the head of NASA’s Goddard Institute for Space Studies, observes, “the five-year mean global temperature has been flat for a decade.”

(Reuters) - Scientists are struggling to explain a slowdown in climate change that has exposed gaps in their understanding and defies a rise in global greenhouse gas emissions.
Here's a little more detail behind what these news articles are popularizing:

Since the slowdown in surface warming over the last 15 years has been a popular topic recently, I thought I would show results for the lower tropospheric temperature (LT) compared to climate models calculated over the same atmospheric layers the satellites sense.

Courtesy of John Christy, and based upon data from the KNMI Climate Explorer, below is a comparison of 44 climate models versus the UAH and RSS satellite observations for global lower tropospheric temperature variations, for the period 1979-2012 from the satellites, and for 1975 – 2025 for the models:

Clearly, there is increasing divergence over the years between the satellite observations (UAH, RSS) and the models. The reasons for the disagreement are not obvious, since there are at least a few possibilities:

1) the real climate system is not as sensitive to increasing CO2 as the models are programmed to be (my preferred explanation)

2) the extra surface heating from more CO2 has been diluted more than expected by increased mixing with cooler, deeper ocean waters (Trenberth’s explanation)

3) increased manmade aerosol pollution is causing a cooling influence, partly mitigating the manmade CO2 warming

If I am correct (explanation #1), then we will continue to see little warming into the future. Additional evidence for lower climate sensitivity in the above plot is the observed response to the 1991 Pinatubo eruption: the temporary temperature dip in 1992-93, and subsequent recovery, is weaker in the observations than in the models. This is exactly what would be predicted with lower climate sensitivity.

On the other hand, if Trenberth is correct (explanation #2), then there should be a period of rapid surface warming that resumes at some point, since the climate system must eventually try to achieve radiative energy equilibrium. Of course, exactly when that might be is unknown.

Explanation #3 (anthropogenic aerosol cooling), while theoretically possible, has always seemed like cheating to me since the magnitude of aerosol cooling is so uncertain it can be invoked in any amount desired to explain the observations. Besides, blaming a lack of warming on humans just seems a little bizarre.

The dark line in the above plot is the 44-model average, and it approximately represents what the IPCC uses for its official best estimate of projected warming. Obviously, there is a substantial disconnect between the models and observations for this statistic.

I find it disingenuous for those who claim that, because not ALL of individual the models disagree with the observations, the models are somehow vindicated. What those pundits fail to mention is that the few models which support weaker warming through 2012 are usually those with lower climate sensitivity.

So, if you are going to claim that the observations support some of the models, and least be honest and admit they support the models that are NOT consistent with the IPCC best estimates of warming.
A couple of notes on the above:

1) In the comments, Dr. Spencer is asked when the models change from hindcast to projection. He notes the physics was frozen in 2007 but was checking to be sure.
2) Note that even in the hindcast portions (i.e. fitted to past behavior), the actual record is always at the lower bound vs. the cloud of model outputs. It is just that now, the actual record is going below even the lower ranges.