The Hidden Cultural Code: How Western-Trained AI Models Export Bias in Non-Western Contexts