The inherent nature of LLMs means their behaviour is probabilistic. All other "jailbreaks" have always worked this way, and have never been explicitly blocked

The inherent nature of LLMs means their behaviour is probabilistic. All other "jailbreaks" have always worked this way, and have never been explicitly blocked by openai (yet) that we know of. Keep retrying until it works. It just did for me.