Comment by eric-burel

You may want to read about agentic AI, you can for instance call an LLM multiple times with different security consideration everytime.