According to the international safety guidelines/standard, the whole-body-averaged specific absorption rate (Poljak et al 2003 IEEE Trans. Electromagn. Compat. 45 141–5) and the peak spatial average SAR are used as metrics for human protection from whole-body and localized exposures, respectively. The IEEE standard (IEEE 2006 IEEE C95.1) indicates that the upper boundary frequency, over which the whole-body-averaged SAR is deemed to be the basic restriction, has been reduced from 6 to 3 GHz, because radio-wave energy is absorbed around the body surface when the frequency is increased. However, no quantitative discussion has been provided to support this description especially from the standpoint of temperature elevation. It is of interest to investigate the maximum temperature elevation in addition to the core temperature even for a whole-body exposure. In the present study, using anatomically based human models, we computed the SAR and the temperature elevation for a plane-wave exposure from 30 MHz to 6 GHz, taking into account the thermoregulatory response. As the primary result, we found that the ratio of the core temperature elevation to the whole-body-averaged SAR is almost frequency independent for frequencies below a few gigahertz; the ratio decreases above this frequency. At frequencies higher than a few gigahertz, core temperature elevation for the same whole-body averaged SAR becomes lower due to heat convection from the skin to air. This lower core temperature elevation is attributable to skin temperature elevation caused by the power absorption around the body surface. Then, core temperature elevation even for whole-body averaged SAR of 4 W kg−1 with the duration of 1 h was at most 0.8 °C, which is smaller than a threshold considered in the safety guidelines/standard. Further, the peak 10 g averaged SAR is correlated with the maximum body temperature elevations without extremities and pinna over the frequencies considered. These findings were confirmed for seven models, including models of a child and a pregnant female. Thus, the current basic restriction for whole-body exposure in the international guidelines is conservative. Peak spatial-averaged SAR can be used as a metric for estimating local temperature elevation even for whole-body exposure. Our computational results also support the description in the IEEE standard about the reduction of the upper applicable frequency of whole-body-averaged SAR from 6 and 3 GHz; the power density reference level is more conservative than the basic restriction limit for the whole-body averaged SAR from the standpoint of temperature elevation.