Individuals with high trait anxiety tend to be worse at flexibly adapting goal-directed behavior to meet changing demands relative to those with low trait anxiety. Past research on anxiety and cognitive flexibility has used tasks that involve overcoming a recently acquired rule, strategy, or response pattern after an abrupt change in task requirements (e.g., choice X led to positive outcomes but now leads to negative outcomes). An important limitation of this research is that many decision making situations require overcoming a preexisting bias (e.g., deciding whether to withdraw a historically winning investment that has experienced recent losses). In the present study we examined whether anxiety differences in the ability to overcome an acquired response extend to the ability to overcome a preexisting bias, when the bias produces objectively disadvantageous decisions. High anxiety (n = 78) and low anxiety participants (n = 76) completed a commonly used measure of cognitive flexibility, reversal learning, and a novel Framed Gambling Task that assessed the extent to which they could make advantageous decisions when the normatively correct choice was inconsistent with a preexisting framing bias. High anxiety participants showed the expected diminished reversal learning performance and also had poorer ability to make advantageous choices that were inconsistent with the framing bias. Worse performance in the Framed Gambling Task was not driven by poor knowledge of risk contingencies, because high anxiety participants reported the same explicit knowledge as low anxiety participants. Instead, the results suggest high anxiety is associated with general deficits in resolving interference from prepotent responses.