In North America, air bag inflation characteristics are determined largely by the unbelted test requirements of US FMVSS 208. Air bag deployment thresholds are set by vehicle manufacturers primarily to prevent facial fracture to unbelted drivers in low speed collisions. This paper examines the extent to which these design practices may be at odds with the protection needs of belted occupants in jurisdictions such as Canada where the seat belt wearing rate is close to 95 percent. Drawing on data compiled in Canada and the US, the field performance of current air bag systems is examined. Analyses of Canadian and US data for both belted drivers (Canada and US) and unbelted drivers (US only) have been performed using an injury/harm model. US injury trends pertaining to belted drivers are in close agreement with the findings of the Canadian field accident studies. Supplementary air bag systems significantly reduce the risk of severe head and facial injuries among belted drivers. However, these benefits are being negated by air bag-induced injuries, most notably to the face in moderate and low speed collisions, and to the upper extremities at all collision severities. Unless current air bag design practices are changed, the widespread use of these systems is unlikely to produce an overall benefit in jurisdictions with high seat belt wearing rates. In particular, these findings require consideration of some specific countermeasures if a satisfactory level of air bag performance is to be achieved in Canada.