Once released into the environment antibiotics can kill or inhibit the growth of bacteria, and in turn potentially have effects on bacterial community structure and ecosystem function. Environmental risk assessment (ERA) seeks to establish protection limits to minimise chemical impacts on the environment, but recent evidence suggests that the current regulatory approaches for ERA for antibiotics may not be adequate for protecting bacteria that have fundamental roles in ecosystem function. In this study we assess the differences in interspecies sensitivity of eight species of cyanobacteria to seven antibiotics (cefazolin, cefotaxime, ampicillin, sufamethazine, sulfadiazine, azithromycin and erythromycin) with three different modes of action. We found that variability in the sensitivity to these antibiotics between species was dependent on the mode of action and varied by up to 70 times for β-lactams. Probabilistic analysis using species sensitivity distributions suggest that the current predicted no effect concentration PNEC for the antibiotics may be either over or under protective of cyanobacteria dependent on the species on which it is based and the mode of action of the antibiotic; the PNECs derived for the macrolide antibiotics were over protective but PNECs for β-lactams were generally under protective. For some geographical locations we identify a significant risk to cyanobacteria populations based upon measured environmental concentrations of selected antibiotics. We conclude that protection limits, as determined according to current regulatory guidance, may not always be protective and might be better derived using SSDs and that including toxicity data for a wider range of (cyano-) bacteria would improve confidence for the ERA of antibiotics.