OpenAI has revealed that its latest artificial intelligence (AI) model, GPT-4o, poses a "medium risk" in its potential to influence political opinions through generated text.