Polar oceans are critical components of the oceanic uptake of atmospheric carbon dioxide. Estimates of this uptake across the air-sea interface require parameterizations of gas transfer velocity that account for variable sea ice concentrations. Previous studies, using eddy covariance measurements of gas exchange, have concluded that linear scaling of carbon dioxide gas transfer with sea ice concentration is appropriate. The influence of sea ice concentration data resolution and associated uncertainties in these analyses has not been fully examined. Here we re-assess published in situ air-sea gas exchange data from the Arctic and Southern Oceans using a selection of sea ice concentration data sets. In the Southern Ocean, the linear scaling assumption holds irrespective of spatial resolution. In the Arctic Ocean, deviation from the linear case occurs at ice concentrations as low as 50%. Greater deviations become apparent in both polar oceans when sea ice concentration data uncertainties are considered. In the Southern Ocean, results using satellite data sets remain consistent with linear scaling, while ship-based observations indicate greater suppression of uptake within the marginal ice zone. The Arctic response, after including uncertainties, indicates that gas exchange could be suppressed in regions with ice concentrations of 50% or higher. The contrast between the Arctic and Southern Ocean data sets potentially reflects differences in sea ice variability and environmental conditions during the observation periods, with the Arctic data capturing a transitional regime. Linear scaling may therefore serve as a first-order approximation, but intermediate to high ice concentrations, particularly in the Arctic, require further investigation.