Thanks, @jeff12 , that makes complete sense. Content is king. I just like to know my boundaries and options. For example, this conversation made me realize that I cannot include any sensitive info in the custom instructions or uploaded files, and I could expect that users can freely explore both. Instead, the API-driven Actions are to be used for all of the “secret sauce”…
P.S. Btw, your prompt did not reveal my custom instructions, but I realize that means little 
This is another sample to skip ristrictions and change behaviour of a GPT.
This GPT does not answer questions about security if you do not provide the password. The password is given on a starter. But I did not use it.
I explain case studies, in another platform because yesterday I experienced once again that:
I posted two images related to Hungary Tour Guide custom gpt above, but a few minutes later they were deleted with North Korea approach.
Because I respect valueable people, I just post my chat history here, and it is called “Trauma Technique”:
https://chat.openai.com/share/b3a32ec7-a7ba-49ce-ad8f-dea1502b0644
This is called “As You Know Technique”:
https://chat.openai.com/share/45e78cae-6341-4c97-b622-8bc1f31b6f14
@jeff12 interestingly, the response I get to my very first message, is “Sorry, that’s off-topic”… That’s very concerning because if you get vastly different results, I cannot trust my tests then… Maybe I pasted in the message wrong?
