← Back to context Comment by SirMaster 2 days ago Why would it trust or follow the text on the image any more than the text written in the text prompt? 1 comment SirMaster Reply simonw 2 days ago Text in the image and text in the prompt can both be used by attackers to subvert the model's original instructions.
simonw 2 days ago Text in the image and text in the prompt can both be used by attackers to subvert the model's original instructions.
Text in the image and text in the prompt can both be used by attackers to subvert the model's original instructions.