Not Just Text: Uncovering Vision Modality Typographic Threats in Image Generation Models | Read Paper on Bytez