The air-water exchange of trace gases such as CO2 is usually parameterized in terms of a gas transfer velocity, which can be derived from direct measurements of the air-sea gas flux. The transfer velocity of poorly soluble gases is driven by near-surface ocean turbulence, which may be enhanced or suppressed by the presence of sea ice. A lack of measurements means that air-sea fluxes in polar regions, where the oceanic sink of CO2 is not well known, are generally estimated using open-ocean transfer velocities scaled by ice fraction. Here, we describe direct determinations of the CO2 gas transfer velocity from eddy covariance flux measurements at a sea-ice lead during the summer-autumn transition in the central Arctic Ocean. CO2 uptake by the lead water is determined using flux footprint analysis of water-atmosphere and ice-atmosphere flux measurements made under conditions (low humidity and high CO2 signal) that minimise errors due to humidity cross-talk. The mean gas transfer velocity over the lead is found to have a quadratic dependence on wind speed: k660 = 0.189 U10^2 which is 25 to 30% lower than commonly used open-ocean parameterizations. As such, current estimates of polar ocean carbon uptake are likely to overestimate gas exchange rates in typical summertime conditions of weak convective turbulence. The gas transfer velocities also exhibit a dependence on the dimension of the lead, via its impact on fetch length and hence sea state. Scaling transfer velocity parameterizations for regional gas exchange estimates will therefore require incorporating lead width data.