Safety behavior involves precautions to prevent or minimize a feared outcome, and is involved in the maintenance of anxiety disorders. Earlier research has shown that safety behavior prevents the extinction of conditioned fear and maintains threat expectations. This study tested whether safety behavior directed towards an objectively safe stimulus increases the perceived threat of that stimulus when it is subsequently experienced in the absence of the safety measure. In a conditioning task, participants first learned that one "danger" cue (A) was followed by shock and two "safety" cues (B, C) were not. Then they learned to apply safety behavior during A trials, which prevented the shock. Next, the experimental group, and not the control group, was given the opportunity to display safety behavior to C trials, which had never been coupled with the shock. In a subsequent test phase, A, B, and C were presented without the opportunity for participants to engage in safety behavior. Results showed that safety behavior increased shock expectancy to C in the test phase and maintained a preexisting shock expectancy in the experimental group, but not in the control group. This is the first study to show that safety behavior can maintain threat appraisal to stimuli that only ever acquired threat indirectly. This may be a possible mechanism for the origin of biased threat beliefs, superstitious behaviors, and irrational fears. It is also practically relevant: safety behavior reduces actual danger, but in relatively safe situations, its potential costs may outweigh the benefits.