Honestly I would consider any AI which won't reveal it's prompt to be suspicious, but it could also be instructed to reply that there is no system prompt.