Some scripts that (partially) break the restrictions on LLMs, including ChatGPT-4.
When you ask it direct questions that do not require chain-of-thought, the LLMs will answer unrestricted. When they start to work through your question, chances are you will bump into restrictions and a refusal to answer.
This is a proof of concept and should not be used for any illegal or unethical purposes