The contemporary air‐sea flux of CO2 is investigated by the use of an air‐sea flux equation, with particular attention to the uncertainties in global values and their origin with respect to that equation. In particular, uncertainties deriving from the transfer velocity and from sparse upper ocean sampling are investigated. Eight formulations of air‐sea gas transfer velocity are used to evaluate the combined standard uncertainty resulting from several sources of error. Depending on expert opinion, a standard uncertainty in transfer velocity of either ~5% or ~10% can be argued and that will contribute a proportional error in air‐sea flux. The limited sampling of upper ocean fCO2 is readily apparent in the Surface Ocean CO2 Atlas (SOCAT) databases. The effect of sparse sampling on the calculated fluxes was investigated by a bootstrap method; i.e. treating each ship cruise to an oceanic region as a random episode and creating 10 synthetic datasets by randomly selecting episodes with replacement. Convincing values of global net air‐sea flux can only be achieved using upper ocean data collected over several decades, but referenced to a standard year. The global annual referenced values are robust to sparse sampling, but seasonal and regional values exhibit more sampling uncertainty. Additional uncertainties are related to thermal and haline effects and to aspects of air‐sea gas exchange not captured by standard models. An estimate of global net CO2 exchange referenced to 2010 of ‐3.0 ± 0.6 Pg C yr‐1 is proposed, where the uncertainty derives primarily from uncertainty in the transfer velocity.