It really depends on how you measure the caffeine. When coffee is roasted the beans lose some water content (somewhere in the 20% range give or take a few percent). At the same time it is losing weight it is gaining size. This leads to a situation that makes answering this question a little interesting.
Assuming all other variables are the same, if you measure by weight you actually have more caffeine in dark roast because the water loss is faster than the minimal caffeine loss during roasting. If you measure by volume you have less caffeine because the beans expand as they roast. This seems to confuse some people so let me restate the above. If you measure your coffee using a scoop you will have less caffeine per cup using dark roast coffee. If you measure your coffee by weight you will have more caffeine per cup using a dark roast. The difference one way or the other is small. If you are buying a cup of coffee and the coffee is measured by weight (common with pre-packed coffee used in many offices and some restaurants) then dark roast will have slightly more caffeine. If you buy a cup and the restaurant measures by volume (common when coffee is fresh ground and measured on the fly) then light roast will be slightly higher in caffeine simply because you will have more coffee grounds. This is really only an issue if you are talking about two identical coffees and even then the differences are small. It is conceivable if you are comparing two available brewed coffees that a difference in varietals between them could make the have as much effect as the roast and the preparation method will almost certainly had a larger effect than the roast level or varietal. If there is a Robusta in one of the coffees it is almost guaranteed to have more caffeine. This is mostly an academic discussion because the differences in caffeine content are relatively small.