A new regression-based copper toxicity model was applied in a case study of San Francisco Bay, California, USA, to demonstrate its utility in estimating risk and site-specific water quality criteria. This was accomplished using probabilistic techniques and a simple model relating dissolved organic carbon (DOC) concentrations with the toxicity of dissolved copper to embryos of the most copper sensitive taxon (Mytilus) in the US Environmental Protection Agency's (USEPA) water quality criteria database. Similar probabilistic techniques were applied to data developed for San Francisco Bay using the USEPA's water-effect ratio (WER) methods for comparison with the DOC-based method. Based on 595 site- and date-specific DOC model observations at 26 sites in San Francisco Bay, none suggested risk of chronic toxicity. Safety factors (1/risk quotient) on average across all sites ranged from 2.4 to 9.1. Comparisons were made between 1) estimates of site-specific criteria made using the DOC method, 2) estimates of site-specific criteria made using the WER method, 3) USEPA national and California Toxics Rule criteria, and 4) region-specific criteria recommended for regulatory implementation by the Clean Estuary Partnership. The DOC- and WER-based methods indicated that copper criteria for San Francisco Bay could be increased above USEPA and California Toxics Rule criteria and will retain the level of protection (≥97%) embodied in the USEPA copper saltwater water quality criteria. The DOC method overall was more conservative (i.e., implies the need for lower criteria in the Bay) than the WER method. The DOC method suggests that the region-specific criteria being recommended for regulatory implementation would be underprotective in some areas and yet could be increased and remain protective in other areas of San Francisco Bay.