Auditing the Ethical Logic of Generative AI Models