University researchers have developed a way to "jailbreak" large language models like Chat-GPT using old-school ASCII art. The technique, aptly named "ArtPrompt," involves crafting an ASCII art ...