We tested the hypothesis that the perceived availability of safety resources plays an influential role in fear prediction bias. Claustrophobic Ss (N = 37) completed a claustrophobic challenge under conditions of either low or high proximity to safety. Proximity to safety was operationalized as distance from the exit. We examined the effects of proximity to safety on Ss' predicted fear, actual fear, and discrepancy between predicted and actual fear (prediction bias). Consistent with prediction, the availability of safety resources had a more pronounced effect on Ss' actual fear than Ss' predicted fear. Moreover, subjects displayed a significant underprediction bias when proximity to safety was low and a slight (albeit nonsignificantly) overprediction bias when proximity to safety was high. Our results lend further support for the hypothesis that anxious Ss underutilize safety information when confronting perceived threats. Possible mechanisms underlying this tendency are discussed.