Summary. There's no factual basis to the claim that there is a 0.1% annual risk of human extinction. What happened is that this figure was used by the authors of the Stern Review as a conservative assumption in their calculation of a discount rate for future costs and benefits, but a game of whispers has caused this context to be lost.
Details. Let's chase down the references and see where this came from, starting with the source given in the OP:
Robinson Meyer (2016). "Human Extinction Isn't That Unlikely". The Atlantic.
The Stern Review, the U.K. government’s premier report on the economics of climate change, assumed a 0.1-percent risk of human extinction every year.
It looks as if Meyer based this sentence on:
Cotton-Barratt et al. (2016). "Global Catastrophic Risks". The Global Priorities Project.
The UK’s Stern Review on the Economics of Climate Change suggested a 0.1% chance of human extinction each year.
Notice that this is not quite the same as the claim in Meyer. Meyer wrote "assumed" but this only says "suggested".
Cotton-Barratt et al. cite this claim to:
Nick Bostrom (2013). "Existential Risk Prevention as Global Priority". Global Policy 4:1 15–31.
The UK’s influential Stern Review on the Economics of Climate Change (2006) used an extinction probability of 0.1 per cent per year in calculating an effective discount rate.
Note that, again, this is quite not the same as the claim in Cotton-Barratt et al. Bostrom says that the Stern review "used an extinction probability ... in calculating an effective discount rate". Using a probability in a technical calculation is not the same as suggesting its truth.
Bostrom cites this to:
Nicholas Stern (2006). "Stern review on the economics of climate change". H.M. Treasury.
Section 2A.2 is a technical discussion of "Intertemporal appraisals and discounting". The problem under discussion is how to compare costs and benefits across long periods of time:
Different strategies for climate change will yield different patterns of consumption over time. We assume that a choice between strategies will depend on their consequences for households now and in the future.
The approach taken is to use a "discount rate" — costs and benefits are accorded lower utility the further they are in the future. The discount rate takes into account various forms of uncertainty about the future, and one of these factors is:
uncertainty about existence of future generations arising from some possible shock which is exogenous to the issues and choices under examination (we used the metaphor of the meteorite).
Some illustrative figures are drawn up (see table 2A.1) to show the effect of various values for the extinction rate δ, and it is at this point that the authors write:
For δ=0.1 per cent there is an almost 10% chance of extinction by the end of a century. That itself seems high – indeed if this were true, and had been true in the past, it would be remarkable that the human race had lasted this long.
It's clear that this figure is just plucked out of the air: no basis is given for it. It's not a serious attempt to estimate the risk of extinction. The point is that unless δ is really high (much bigger than 0.1%) it is dominated by the other contributions to the discount rate in the model (namely elasticity of marginal utility and growth rate) and so the Stern Review's conclusions don't depend on its exact value, and so there is no need for them to make a serious estimation.
Update. After the problems with the Atlantic article (see this article by nostalgebraist) were brought to their attention, the Global Priorities Project issued an erratum:
We were aware that the Stern Review used this figure merely as a modelling assumption, and were trying to give a concise accurate statement. Our intention in using the figure from the Stern Review was not to try to pin down an accurate estimate of the likelihood of global catastrophe, but to demonstrate that existing serious analysis treats the 0.1% probability as a plausible modeling assumption, which would have consequences that are interesting and non-intuitive. [...]
The car crash comparison was picked up in The Atlantic, which reported it as an unconditional claim and emphasised it in their article. We did not intend to argue that the 0.1% figure was an accurate estimate of extinction risk (as we did not plan to offer an estimate of extinction risk), so this was inadvertently misleading to Atlantic readers.