Even with these in place, the output of the LLM is not trustworthy. In the sense that it doesn't care what it true, only what is plausible.