By focusing on the skills that large language models lack, ‘designgineers’ can adapt to a market upended by AI.

The exploit, known as a prompt injection attack, evades detection by reducing the prompt font size and changing it to white to blend in.