Researchers were able to jailbreak five state-of-the-art (SOTA) large language models (LLMs), including Open AI’s ChatGPT-3.5, GPT-4, Gemini, Claude, and Meta’s Llama2 using ASCII art.

I'll dust off my old ascii art of naked women from my hornier youth and see if I can make an AI pain a hyper-realistic depiction of such, you know, for science! More options Report Add bookmark Share ...

LLMs are trained to block harmful responses. Old-school images can override those rules. See full article ...