This is a very interesting idea!
As you stated above, the biggest question might be "which religion" should we give these AI models, and the impact of our choice on the rest of humanity. (Optimistically), I think that the answer to this might be simpler: let the model decide for itself what is true. If there really is sufficient epistemic evidence that what you believe is true, (which one might assume you believe as you are Jewish, although I am Catholic) then a sufficiently intelligent (e.g. one powerful enough to cause us worry) model should be able to ... (read more)