Utilizing a robust mean–variance model that incorporates smooth ambiguity preferences, we derive a closed-form solution for the optimal currency exposure. Within this theoretical framework, the demand for optimal currency hedging is formulated as the solution to a generalized ridge regression. Our findings indicate that the investor's aversion to model uncertainty increases the demand for hedging. The empirical analysis illustrates that ambiguity introduces greater estimation bias and narrows the confidence interval of the optimal currency exposure estimator. An out-of-sample backtest further demonstrates that incorporating ambiguity into the model improves the stability of optimal currency allocation over time and significantly reduces portfolio volatility after accounting for transaction costs.