Safespeed believe that there has been a trend for speed limits to be lowered without evidence that lowering them will reduce crashes. They believe that speed limits "should never be used to attempt to modify the speed of traffic" and that the claim that "accident risk simply increases with speed...is false". However, they do "welcome properly set speed limits, and welcome speed limit enforcement when speed limits are exceeded in a way that causes danger". (source) Those statements don't sit well together. If you say "you should never attempt to modify the speed of traffic using speed limits", then how would you slow traffic if it is fast enough to cause danger? Their Road Safety Manifesto should give us a clue: it concentrates on "driver training", "road safety culture", and "road safety information should be regularly aired on TV at peak viewing times". Their Manifesto doesn't cite any evidence that any of this will work. It's all supposition and wishful thinking.
Meanwhile, back in the real world:
There is some interesting Dutch research here. It studies the effect of reducing motorway speed limits from 120kph or 100kph to 80kph, accompanied by strict enforcement. The reason it is interesting is that speed limits were reduced not to reduce crashes, but to comply with air quality legislation. So the 'regression-to-mean' effect that Safespeed usually use to cast doubt on the correlation between lower speeds and lower crash rates does not apply. The study says "The effects at A13 Overschie were spectacular (empirical results): the number of accidents decreased by 60% and the number of victims by 90%. Besides, no indications were found of compensatory behaviour (drivers who speed up downstream). The increased safety was caused by lower speeds and lower speed differences."
Futher research: Nilsson (1982) researched speed limit changes from 110 to 90 km/h on Swedish roads. He compared changes in accident risks with roads where the speed limits had remained unchanged, and reported the statistical relations (the regression model) of speed (limits) and accidents. This model is now well-known as the Nilsson formulas. You can read about it here.
But is speed in itself correlated with increased crash risk? According to the EU's road safety experts, "A higher speed increases the likelihood of an accident. Very strong relationships have been established between speed and accident risk: The general relationship holds for all speeds and all roads, but the rate of increase in accident risk varies with initial speed level and road type. Large speed differences at a road also increase the likelihood of an accident. In addition, drivers driving much faster than the average driver have a higher accident risk; it is not yet evident that this is also the case for the slower driver."