In my experience, ChatGPT was perfectly happy providing logic for a question that counteracted previous logic it provided. When I pointed out this conflict, it disagreed that it was in conflict when it clearly was.