Abnormal fear responding to threat cues may contribute to the aetiology and maintenance of persistent fears and pathological anxiety. Chronic anxiety may also involve abnormal fear responding to ‘safety’ cues, which do not signal danger. Yet investigations of fear responding to acquired safety cues are scarce and the basis of such responding remains unclear. Moreover, previous studies do not distinguish between stimulus generalization (an associative mechanism based on perceptual similarity between threat and safety cues) and sensitization (a non-associative mechanism whereby fear responses to any novel, intense, or fear-related stimulus are temporarily elevated). This study investigated responses to acquired safety cues in volunteers with varying trait anxiety, using a novel fear conditioning paradigm designed to distinguish between effects of trait anxiety on generalization and sensitization. The paradigm used three conditioned stimuli: a threat cue (CS+) and two safety cues (CS−), one perceptually similar to the CS+ and one perceptually dissimilar. Conditioned fear to these cues was indexed by fear potentiation of the startle blink reflex, skin conductance responses, and self-report. To examine how trait anxiety moderated responses to safety cues, participants were divided into high and low trait anxiety subgroups. Startle, skin conductance, and self-reported fear measures indicated that generalization of fear occurred for the safety cue which resembled the threat cue, but not for the perceptually dissimilar safety cue, consistent with the stimulus generalization hypothesis. There was some evidence that stimulus generalization was exaggerated in anxious individuals. The current study sheds light on the mechanism by which fear responses to safety cues arise in healthy individuals, and offers some insight into the influence of this mechanism in chronic anxiety.