“For users to form healthy relationships with AI systems, it is important that the models are transparent about their nature, such as how they were trained, and do not misrepresent themselves.”