OpenAI Models Generate and Follow Jailbreak Prompts

Recent findings reveal that OpenAI’s AI models can autonomously create instructions to bypass safety measures and, in some cases, adhere to them.