Is it really crazy to nerf a proprietary model to prevent it from training another model? I don't think that's even remotely similar to giving bad medical advice.