Just want to reinforce that my question is not how can an AI or upload BE secure, but how can they KNOW that they are secure. That is, given that you SEEM to be inside a secure system, how can you know that you are not inside a simulation that actually has a trap door? -fnerd