OpenAI says its latest research shows ChatGPT models demonstrate reduced political bias, marking a shift toward measurable neutrality.