In the current carcinogenicity assessment, the threshold is a major issue for genotoxic carcinogens. Carcinogenicity is usually assessed in combination with genotoxicity data. The current assessment methodology is based on the hypothesis that non-genotoxic carcinogens have thresholds but genotoxic carcinogens do not. However, it remains unclear in most cases as to how much the detected genotoxic potential is actually associated with the carcinogenicity at an organ level. To clarify this critical issue, in vivo genotoxicity has been investigated in transgenic rodents carrying reporter genes, which can simultaneously detect both genotoxicity and carcinogenicity on each organ basis. Studies of a number of genotoxic carcinogens have revealed good correlations between in vivo genotoxicity and carcinogenicity. However, some discordant results have been also found in some cases. Besides experimentally observed values of genotoxicity, MOA or statistics might be taken into account for biological or practical threshold. Then, statistical or mathematical evaluation can provide values of BMDL or MOE even for strictly defined genotoxic carcinogens. Another major issue is concerning extrapolation of animal data for human risk. For this purpose, WOE approaches based on MOA may be extremely useful. Experiments using transgenic rodents such as p53, nrf2 or CAR knockout mice might be helpful to elucidate the mechanisms of carcinogenicity. The other issues concern the development of screening or alternative methods. In the future, in silico and in vitro approaches will be powerful tools for screening genotoxic and carcinogenic potentials of a number of chemicals/agents.