Using a competitor's AI model outputs to train a rival system is ethically acceptable